Download as pdf or txt
Download as pdf or txt
You are on page 1of 40

On 'Thurneysen-Havet's Law' in Latin and Italic

Author(s): Brent Vine


Source: Historische Sprachforschung / Historical Linguistics, Bd. 119 (2006), pp. 211-249
Published by: Vandenhoeck & Ruprecht (GmbH & Co. KG)
Stable URL: http://www.jstor.org/stable/40849384 .
Accessed: 17/06/2014 14:46

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Vandenhoeck & Ruprecht (GmbH & Co. KG) is collaborating with JSTOR to digitize, preserve and extend
access to Historische Sprachforschung / Historical Linguistics.

http://www.jstor.org

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
On 'Thurneysen-Havet'sLaw' in Latin and Italic

1. Umbriansauitu and relatedmatters


1.1. In a recentpaper (2004:622ff.),I proposedan analysisof Um-
briansauituthatappealedto theoperation,withinthehistoryof Umbri-
an (or earlier),of "Thurneysen-Havet's Law", thephonologicalprocess
thataccounts forthe a-vocalism of Latin formslike caueõ 'bewareof
Since thisis a processforwhich
[: Koéo) 'perceive'] < *kouh1-éio/e-.ì
non-Latinexampleshad not hithertobeen noted,the Umbriandatum
has interesting consequencesfor Italic phonologythatdeserve to be
exploredfurther. I begin by brieflyrecapitulatingthe analysisof the
Umbrianform,together withotherintroductory material.
1.2. The famouscurseagainsttheenemiesof theUmbrianstate(Igu-
vine Tables VIb 60, Vila 49) consistsof fivealliteratingpairsof 3 sg.
imperative ("futureimperative")forms;in theVIb 60 version:

[.] hondu. holtu. ninctu. nepitu. sonitu. sauitu .


tursitu . tremitu
.
preplotatu preuilatu

In context,a trio of divinitiesis asked to performthe actions


expressedby theseverbs:"[They] shall terrify [them],cause [them]to
tremble,..." etc.2As is for
typical Umbrian, severalof thelexical items
here are etymologically difficult,
includingsauitu? Still, the formis
widelyconjectured to have some with
relationship Lat. saucius 'wound-
ed' (thusvon Planta,Buck, Devoto, Pisani, Olzscha, Poultney,Ancil-
amongothers),e.g. via a pre-Latin*sau-ikio-'and although
lotti-Cerri,
the backgroundof saucius itself [Liv. Andr.+] is entirelyunclear

1The
phoneticnatureof the process will be discussed below; forthe moment,a de-
scriptivecharacterizationsuch as "pre-Latin*(-)ouV- > Latin (-)auV-" will suffice.À
propos the conventionalname: as pointed out by Horton-Smith(1895: 457), it was
Saussure, in his Mémoire,who firsttook note of the process as a systematicpheno-
menon(1879: 104).
2 On the of the singularverbformswithpluralsubject(vs. Vila
syntacticirregularity
49, witha singledivinityas subject),see myearliertreatment (2004: 615), withfurther
references.
3 "B[e]d[eutung]unbekannt... Etimologie] ohne B[e]d[eutung]-Bestimmung ungesi-
chert"(Untermann 2000: 658, q.v. foradditionalbibliographicaldetail,includingmate-
rialcitedbelow by author'snameonly).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
212 BrentVine

(Walde-Hofmanns.v.: "unerkl[ärt]"),this formallyand semantically


attractive comparisonoffersby farthemostpromisingapproachtoward
interpreting U. sauitu.An important stepwas takenby Ribezzo (1931),
who comparedtheformwithVed. ksnu-'whet' (RV ksnaumi)and Gk.
Çwû'scrape', an immediately appealingcomparisonfora word(or, in-
cluding Lat. saucius, words) that may mean 'wound' or the like. But
seriousdifficulties are posed by the rootvocalismijust as thereis no
PIE "*mh-"thatcan underlieLat. saucius,thereis no "*ksau-"either-
butthereis good evidence,as Ribezzo saw, fora verbalroot*kseu-(see
LIV s.v.): in additionto Çiko,note also Ved. ksurá- 'razor' = £opóv
'id.', and (despiteLIV, whichsetsup a separateroot*ksneu-)probably
also original «-infix forms(in addition to RV ksnaumi also Lat.
nouãcula 'razor').4The problemmaythusbe restatedas follows:how
can one accountin a naturalway forthetransition froman old *ksou-
éio/e-(presumablyiterative,'scrape/woundagain and again') to the
Italic *(k)saueio/e-thatcan underlieU. sauitul My suggestionwas that
"Thurneysen-Havet's > Lat. cauëre etc.)
Law" (cf. 1.1., *kouhj-éio/e-
providesan immediateand attractive solution,despitethefactthatthis
processhad hitherto been establishedonly forLatin.5There is, never-
theless,good reason to suspectthatthis process could have operated
outsideLatin,given its relativechronology(as discussedin detail by
Schrijver):it mustprecedenotonly *VguVand *VgyhV > *VuV,which
mustbelongto a veryearlystratum of Latin (hencefoueõ 'warm',not
**faueõ'nüdus 'naked',6not *naudus,etc.),but it also precedes*eu >
*ou (thusnouas 'new', not **nauus,etc.),7whichis oftenassumedto
be an innovationof Proto-Italiedate, and whichin any event(thatis,
despitethepossible survivalof a few -eu-formsintoearlyLatin) must
be consideredto be veryold.8For thesereasons,Schrijverconcluded

4Ontheentire complex,seeGarcíaRamón2000: 121(including enlarged *ks-en-and


*ks-es-besideour*ks-eu-, to *kes-'ordnen':OCS cesati'comb'etc.).
originally
5 Forfulldiscussionof Thurneysen-Havet'sLaw (withearlierreferences),see Col-
linge1985: 193ff.,Schrijver1991:436ff.(on whichsee Rasmussen1993:200), and
Meiser1998: 85. (Earlierhandbooktreatments can be foundin Sommer1914: 109,
Sommer-Pfister Leumann1977:49f.)(See note57 belowforthetreatment
1997:159ff.,
bySihlerl995.)
6 Assuming an o-gradesourceofthetraditional
type*nogu-Vd o- (butsee Nussbaum
1999:383,401).
7We will,however, haveoccasiontoreconsidernouusbelow(3.5.).
8 in question(esp.neven
Despitee.g.Leumann1977:70f.,all oftheLat.-eu-forms
CIL I2 455 [Ardea],nevna CIL I2 2845-6[TorTignosa],Carm.Sal. Leucesie)have
plausiblesecondary anddiscussion
references
explanations; atVine1993:63n56.As for
theprobablesecondary /eu/diphthongs in thedialectalLatinforms, parallelsare not
to find:BruceHayespointsoutto methe[ow] > [ew]development
difficult in some

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 213

his analysisof thechronology of Thurneysen-Havet's Law byobserving


(1991: 453) thattheprocess"can in principlebe veryold" and that"No
othertypicallyLatin developmentcan as yetbe dated beforeit"; and
Meiser (loc. cit. [n. 5]) arrivedat thesame conclusion,assertingthatin
theory,the rule is "auch fürdas Sabellische gültig"- as indeed von
Plantahad alreadysuspected(1892: 115), withoutbeingable to provide
any clear-cutexamplesin Sabellian. Such a view would be confirmed
by U. sauitu,if it is to be explainedon thebasis of (a pre-Latinversion
of) Thurneysen-Havet's Law; and an attractive case fromVeneticwill
be presentedbelow (4.4.1.)- It is preciselytheprobableantiquity of the
processthatis central to thediscussion in sections 2. and following.
1.3. Two possiblecounterexamples in Sabellianare knownto me.The
firstis U. anouihimu(VIb 49, bis; 3 sg. mid.imperative),9 witha prob-
able meaning 'put on (clothing/insignia etc.)' (cf. Lat. induõ, Arm.
aganim 'id.', etc.; LIV s.v. *h2euH-,Untermann 2000: 112f.).Although
theformis ambiguousin termsof its inflectional class (it can continue
eitheran ë-verbor an f-verb), it could well reflectan iterative-causative
*-h2ouH-éio/e-, which mighthave been expectedto surface,withan
applicationof Thurneysen-Havet's Law, as ^anauihimu.This is not,
however,a probativecounterexample,since the vowel in question
appearsin a medialsyllable(afterthepreverb).Thatbeingthecase, the
o-vocalismcould be explained,as suggestedby Meiser(2003: 67nl 13),
withrecourseto themedial-syllable a > o roundingprocess(withadja-
centlabial) alreadyestablishedforUmbrian(prehub- beside prehab-,
comoltubeside kumaltu,etc.; Meiser 1986: 268ff.).Nor is originalo-
vocalisma necessarystarting pointin any case: theformcould just as
easily continue an e-grade io/e-verb (*-h2éuH-io/e- > pre-U.*-auio/e-),
as Meiserassumes.10
U. -tu'itu/-douitu, -tuvies(together withFai. douiad and OLat. duim,
duat) will be in
addressed 4.4.7. below.
1.4. Finally,it is worthcommenting brieflyon some "non-evidence"
forThurneysen-Havet's Law in Umbrian,despitehandbook(and other)
treatments. CertainUmbrianpreformsare oftenwrittenas if Thurn-
eysen-Havet'sLaw had operatedat some pre-Umbrian stage: thusU.
kutef,traditionally interpreted to mean 'tacitus' vel sim., as if equi-

versionsof early20th-century BritishReceived Pronunciation(Wells 1982: 293f.), as


well as thesame phenomenonin some Americandialects.(No -eu-formsare attestedin
Faliscan or Sabellian.)
9 I am to Michael Weiss forbringingthisformto myattention.
grateful
10
Similarlyvon Planta,Brugmann,Buck and others,see Untermannloc. cit. (where
also forstillotherderivationalpossibilities,such as denominativeformation).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
214 BrentVine

valentto a Lat. *cautëns(and witha formation like Lat. fa-t-ë-rï,cf.


Ose. fatíum'assert'),is routinely assigneda preform*kau(i)-t-ë-or the
like, as if showinga pre-Umbrian applicationof Thurneysen-Havet's
Law (see e.g. Untermann2000: 422). n But in all such forms,the
surfacemonophthong in Umbrianis phonologicallyambiguous(thus
here <u> in principlefromPr.-Ital. *<zwor *ow), and provides no
information at all abouttheearliervowelquality.12

2. On theapparent Latin exceptionsto Thurneysen-Havet'sLaw


2.1. If Thurneysen-Havet's Law operatedat a veryearlyperiod (as
had alreadybeen suspected),it becomes important to reexaminepre-
viously-identified exceptions - all of them, course,in Latin. These
of
have been analyzedextensivelyby Schrijver(cf. n. 5), whosetreatment
lefta smallbut significant residueof unexplained,ambiguous,or (as it
seemsto me) unsatisfactorily explainedcases. (We will also consideran
additionalexceptionalform,to myknowledgenotpreviouslydiscussed
in thisconnection.)
2.2. The firstsubtypeto be evaluatedinvolvesLatin formswiththe
descriptiveappearance of containingan "unshifted"prevocalic -ou-
diphthong, forwhichthe simplestexplanationwould appeal to old e-
grades (thereforenot subject to Thurneysen-Havet's Law, cf. 1.2.
nouus); thusSchrijver(1991: 451): "eu ratherthanou is possiblefor2.
cloaca, 7. mouëreand 8. nouãcula". For such explanationsto be con-
vincing,it is crucial to providemorphologicaljustificationforthe e-
grades.For two of these items(nouãcula and cloaca), Schrijverpro-
vides no actual explanation,and his account of the third(mouëre)
seemsproblematic. A fourthitem(an alleged Lat. ^touëre),nottreated
by Schrijver, calls fordiscussionas well.
understoodas an
2.2.1. Lat. nouãcula 'razor' [Cic.+] is traditionally
instrument noun to a verb *nouãre (cf. gubernãre'steer' - ►guber-
nãculum'rudder',etc.), ultimately belongingwiththe root *ksneu-of
RV ksnaumi (1.2.); but the source of this *nouãre is variously
described,or sometimesleftunspecified.Thus,Ernout-Meillet provide
a reconstruction *ksnouã-tlo- (sic), which(withits pre-Italic*ksn-and
*-r/o-)seems to assume an inheritedö-grade;in contrast,Walde-Hof-
manncite Kretschmer (1892) as one of theauthoritiesfortheire-grade
*qsneuã - and yet Kretschmerhimselfwas non-committal on this

11More Untermann 2002:379nlO("umbr.kutef< *kau-t-ë-ns").


recently,
i¿Likewise (see Unter-
U.fons 'fauens, oftenassumedtocontinue
Sab. *fau(V)-ni-
mann2000: 302), butin factmultiplyambiguous, has discussed(1991:
as Schrijver
442).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 215

point.13 As alreadynoted,Schrijver(1991: 449) pointsout theapparent


Thurneysen-Havet's Law violation and the formalambiguity("Lat.
nouãcula reflects*ksneu-or *ksnou-"),withoutfurther comment.
Whatsortof morphological basis mightan e-gradeformation have in
thiscase? The mostimmediatepossibility- a primary e-gradeabstract/
verbalnoun *ksneu-eh2- (> pre-Lat.*ksneuã-)- is notat all attractive:
no examples of such a type are foundin Latin at all, as shown by
Livingston(1997: 64) in heranalysisof Lat. praeda 'booty' (besttaken
froman o-grade *prai-hod-ã,ratherthanthe conventionally-assumed
*prai-hed-ã).Nor are such e-gradeabstractswell-attested elsewhere;
thus,the only formin Greek witha claim to archaismis Horn.Ei)%í'
'prayer,vow', whichis bettertakenas déverbalto si^oum 'pray,vow',
accordingto a productive pattern withinGreek.14
As a preliminary to an alternativesuggestion,we must address a
semanticquestionthatseems generallyto be ignored:if PIE *ksneu-
meant 'whet,sharpen' (its traditionalgloss, cf. RV ksnaumiand its
instrument noun ksnótra-'whetstone'),whydoes the instrument noun
Lat. nouãcula mean 'razor' (and notsomething morelike 'whetstone')?
The simplestassumptionis thattheinfixed"pseudo-root"*ksneu-(1.2.)
is betterconceivedas an iterativeversionof *kseu-,witha meaninglike
'scrape (repeatedly)'- i.e., untilsharp(as in Indie) or untilsmooth(as
in Italic).15
Thus,as a possiblesolutionto theThurneysen-Havet's Law problem,
I suggestthatwhat may underlieLat. nouãcula is a primarye-grade
thematicadjective *ksneu-o-'(scraped) smooth' of the type *néu-o-
'new', *sén-o-'old', and severalothersuch forms.That thistypecan
also be built on verbal roots is evidentfromexamples like Heuk-ó-
'bright'(LIV *leuk-'hell werden'),*leubh-o-'dear' (LIV *leubh~'heb
sein etc.').16The more specific morphologicalclaim, then,is that a

13 "Lat. noua-cula ist von einem verloren


gegangenen verbum Lat. *nouã-re
'scheeren' ausgegangen, welchem seinerseitswieder älteres *ksneu-mizu gründe
gelegenhaben wird"(1892: 470); butcf. 1892: 419 on RV ksnaumi:"... danebenidg.
snov- oder snev- [emph.BV] in lat.nou-ã-cula 'schermesser'aus *snouãcula".
" Ct.
auDxn tearing,scratching<- auDaaco tear,scratch, kouiòti care <- kouiÇcû
'takecare of, etc.; Chantraine1933: 23, Peters1980: 12f.
n On the iterativevalue of RV ksnaumi,see García Ramón 2000
(with the reser-
vationsexpressedby Klein,2003: 46).
1ULike ^leuk-o- - ►àeukoç whitehsh, ÀeDKT|(whitish)skincondition;white
(AevKOÇ
poplar') probablyalso *per£-ó-'striped,speckled' (*7repKÓç- ►rcépKOÇ 'hawk', 7iépKr|
'perch') and *dheudh-ó- 'yellow,brown' - >
(*t£D$óç xeu&oç 'squid'; thusJ. Schindler,
1978 lectures,cf. C. de Lamberterie,CEG 6 [2001] 151). Note also *hjreudh-o-'red',
*h2erg-o-'bright'(see Rau 1998: 159 forthelatter,vs. traditional*ti2rg-[r]ó-
> ápyóç,
withdissimilation).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
216 BrentVine

*ksneu-o-'(scraped) smooth'servedas thebasis fora factitivein -eh2-


*ksneu-eh2-(io/e-) 'to makesmooth(by scraping)',whencea Latinverb
*nouãreand itsinstrument nounnouãcula 'razor'.17
As an alternative,one could consider a Latin denominativeverb
*(ks)neuãio/e- based on an e-gradeneuternoun *ksnéu-o-,of thetype
seen in *uérg-o-'work' and a seriesof otherforms,in generalquiteiso-
lated.18At bottom,a principledchoice betweenthefactitiveand deno-
minativeoptions(bothbased on ^-grades)maynotbe possible.
2.2.2. The situationwith clo(u)ãca 'sewer'19 is in many ways
comparableto thatof nouãcula (cf. Schrijver:"cloaca reflects*lcleuH-
or ^louH" [1991: 448]). The generally-accepted Baltic cognate(Li.
[Zem.] slavu 'sweep, clean') has the appearance a primaryö-grade
of
present, and is thus notimmediately helpful,especiallyifone wishesto
support an old e-grade fortheLatin form. Thereseem,in fact,to be two
main views about the sourceof cloaca:
(i) it is based on an abstractnoun *clouã- 'Reinigung'(e.g. Rix 1999:
519);
(ii) it is deverbativeto a *clouãre 'reinigen'(e.g. Leumann 1977:
340);
and (thoughnotedneitherby Rix norLeumann)bothsourceswould
violate Thurneysen-Havet's Law if the word has an old *-<?w-di-
phthong.To complicatematters, a Lat. cloãre 'reinigen'actuallyexists
- attestedonce, in a gloss (Serv. auct. ad V. A. 1.720, uetereschare
purgare dixerunt,explaining the epithet Cloacina). This gloss,
moreover,is suspiciouslysimilarto the sole attestation of a synony-
mous verb cluere (Plin. Nat. 15.119, cluere enim antiqui purgare
dicebant,explainingCluacina). Rix explainscloãre on thebasis of an
old ö-gradecausative(again withno noticeof theThurneysen-Havet's
Law violation); but he properlyacknowledgesthatcloãre mightbe
nothing more than "eine Erfindungvon Servius' Quelle (wohl
Varro)".20 We maythusreturn to cloaca itself.
Of thetwoviewsoutlinedabove aboutthesourceof cloaca:

17Such
(of thetype*néu-o-'new' - ►renouãre'renew') are productive
-e/i2-factitives
in Latin. On the type,see recentlyJasanoff2003: 139ff.,Tremblay2003: 153nl77; for
therelatedAnatolianmaterial,see Melchert1997: 132ff.
18Thus also
*péd-o- tracketc. , *jeu-o- barley, and einzelsprachlichitems: Lat.
serum 'whey', OPr. kelan 'wheel', Hitt.egan 'ice', and others.(PIE *uérdh-o-'word'
[Lat. uerbumetc.] is oftenplaced in thiscategory,but mayinsteadcontinuean original
compound*uéríh¡]-dhh¡-o-, as discussedby Hackstein[2002: 14].)
19See Leumann
(1977: 99) on theLatin spellingvariation(clovaca, cluaca).
2USimilar
skepticismabouttheverbis expressedby Kümmel(LIV s.v. *kleuH-).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' inLatinandItalic
On 'Thurneysen-Havet's 217

(i) as noted, an old o-grade abstract*idou-eh2-> ^cloua- would


violate Thurneysen-Havet's Law; and an e-grade abstract*fcleu-eh2-
facesthesameobjectionsas *ksneu-eh2- (2.2.1.);
(ii) in contrast,a Lat. *clouãre(withfactitivemeaning)- possiblythe
same chare thatis actuallyattested- wouldbe parallelto *(ks)nouãre;
thatis, an originale-gradethematicadjective ^ideuH-o-'clear, clean'
would serveas thebasis fora factitive*£leuH-eh2-(io/e-) 'make clean',
neitherof theseformsbeingsusceptibleto Thurneysen-Havet's Law.
the
Finally,although point cannot be pursuedhere, note thatthisroot
makesa primaryzero-gradero-adjective,as attestedin Celtic (W. dir
'clean' < ^IcluH-ro-), beside thee-gradethematicadjectivejust posited
- a patternthatcan be paralleledforothersuch^-gradeadjectives;thus
"ideuH-o-: *ftluH-ró- like *h}reudh-o- : *hirudh-ró- (épuopóç 'red')
and ^h2erg-o-: ^h^g-ró- (Ved. rjrá- 'quick'), to whichmighteven be
added *kseu-o-: *ksu-ró-(ksurá-,£opóv,1.2.), ifitcould be shownthat
"*ksneu-o-"(2.2.1.) had replaced an earlier *kseu-o-(among other
uncertainties thiscase).
affecting
2.2.3. Superficially ratherdifferent (thoughultimately similar)is Lat.
moueõ,universallyregardedas continuingan old o-gradecausative,21
even thoughtheblatantviolationof Thurneysen-Havet's Law thatsuch
a preformwould entail(cf. Lat. caueõ, 1.1.) is all but ignoredin the
morphologicaland etymologicalliterature. An exceptionis G. Meiser,
whoprovidestwosuggestionsforexplainingtheviolation(1998: 85):
(i) eithertheloi has beenrestoredafterthecausativefoueõ 'to warm';
(ii) or theunrounding of theloi in moueõ was blockedphonetically,
its
given position between two labial segments(similarly,accordingto
for
Meiser, bos, bouem, anotherapparentThurneysen-Havet's Law vio-
lation, to be addressedseparatelybelow).
Ad (i): the formalresemblancebetweenmoueõ andfoueõ may well
have been a factor(as will also appear in anotherconnection,2.2.4.);
but thisis not likelyto have been a sufficient cause, especiallysince
moueõis partlyintransitive (moreon thisbelow), and thusfunctionally
distinctfromcanonical causatives like foueõ. As for the phonetic
explanation(ii): in its favor is the fact thatLatin otherwiseattests
certainroundingprocessesthatoperatein similarenvironments, as in
thetypeuomo 'vomit'< *uemõ.22

21ThusErnout-Meillet,
Walde-Hofmann, Leumann1977: 541, LIV s.v. *m}euh¡-
(Zehnder,
Kümmel).
11The rulethat
governsthislast item,however,
is rathermorecomplexthanis
generally cf.Nussbaum
thought, 2003.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
218 Brent
Vine

More attentionhas been devoted to the moueõ problem in the


literature on Thurneysen-Havet's Law.23Schrijvertentatively suggests
thatmoueõ continuesan "e-gradecausative" (*mieuhréio/e-),2A com-
paring Lat. augëre increase (trans.)',suãdere 'urge', terrëre 'frighten',
cëuëre 'wiggle (the hips)', and merëre'be deserving'.But this is a
questionabletype,as Schrijverhimselfseemsto realize,whenhe writes
that"theformation mightnotgo back to PIE."; indeed,all of thecases
he names can involve secondarydevelopmentsor otherexplanations:
thus,the ¿-gradeof augëre could be based on its s-aorist(auxi) (thus
LIV); thevocalismof suãdere ('urge' < *'sweeten') could be based on
*sueh2d-u-'sweet' (cf. suãuis) (thus LIV); terrërecan be phonolo-
gicallyregular< Hros-éio/e-(Nussbaum 1999: 412n87, LIV) or with
vocalismafterterror(< Hrés-õs)(Nussbaum);merëre/ -I 'be deserving'
may ratherbe a stativein *-ehr (similarlyLIV, Meiser 2003: 91);
cëuëre has a perfect"cëur attestedonly in grammarians,where
evidenceis also foundforthematicinflectionas cëuëre (see TLL III:
982.29ff.),and the sole cognateappearsto be CS po-kyjç 'shake (the
head)' (cf. LIV s.v. *kehju-),all of whichmeansthatthereis too little
information available to make a determination as to themorphological
basis of cëuëre.
There are, however,othermorphologicalpossibilitiesthatmay help
accountformoueõ;butgivenwell-knowncomplexitiessurrounding the
shape of the rootitself(conventionally *mieuhr/*mihju-, as in LIV),25
as well as difficultieswithsome of thealleged cognates(especiallyin
Anatolian26), thesecan onlybe sketchedhere.Still,sucha sketchshould
suffice to demonstratethat there are reasonable possibilities for
explainingLat. moueõ on the basis of one or more inherited¿-grade
formations.
The startingpointforone such approachis an observationalready
regardedby Schrijveras potentiallyimportant (althoughhe does not
the
pursue matter): he notesthat,given theintransitive usage of moueõ,
theverb"cannotbe considereda typicalcausative"(1991: 449). Indeed,
intransitive behaviorfor moueõ is extremelywell-attested(fromthe

23 In recenttimes,note
especially Collinge 1985: 194 and, with an extendeddis-
cussion,Schrijver1991: 448f.,469.
24Thus 1895: 452nl ("moueõ < *meueõ,cf.áusúaao&ai").
alreadyHorton-Smith
25Or to whom
as suggestedto me by JayJasanoff,
perhapsrather^miehi-u-ñmihi-u-,
I am indebtedforvaluablediscussionaboutthematerialin thissection.
26I thank
Craig Melchertforhelpfulguidanceaboutsome of theAnatolianmaterialin
thissection.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 219

Archaicperiodon),27mostfrequently, in the infectum,forthe present


participleand infinitive, and moregenerallyforthe perfect(the latter
includingthe idiomaticusage fordescribingearthquakes:terramouit
etc.). I suggest,then,thatifthisbehavioris old, Lat. moueõcould result
froma conflationof two distinctinheritedformations, one transitive
(the actual causative) and anotherintransitive.28 What is normally
reconstructed forthe PIE verb is a root athematicpresent(e.g. LIV:
*miéuhrti/*mihiu-énti -» thematizedVed. mïvati'pushes'), thuswith
in
¿-grade strong forms - a formationthatcould be continued,directly
or indirectly, in the Latin intransitive usages. Formally,intransitive
mouëns(to a thematic*mouere)could continuea thematized*miéuhr
e-ti29or (moredirectly)could resultfroman athematic*miéuhrti(with
developmentas in uomere <- *uemhrti).30 Similarly,an old ¿-grade
intransitive rootaorist*miéuhrt(withregularreplacement in Latin by
M-perfect *meua-uai) could have ended up directlyas the (intrans.)
perfectmõui,beside the ¿-gradeverbal noun in *-iw-*miéuhrtu-(>
supine mõtus 'movement,motion'). Thus, froma historicalpoint of
view, the Latin second-conjugation stem"moue-"(withsynchronieo-
vocalism)could reflecta compromisebetweenthestem-formation of an
originalcausativeand thevocalism(originallyinvolvingone or moree-
grades)of the inheritedintransitive.In such a development, the voice-
neutralsupine may have played a role (note expressionsof the sort
terraemotus'earthquake',cf. terramouitabove), as well as theformal
overlap- apartfromthe Thurneysen-Havet's Law treatment - in the
Latinpresentparticiplesof botha plainthematicverb(intrans, mouent-)
and an originalcausative(^mauent-).
A scenariolike thatjust outlinedwould seem to be unnecessary,in
view of the standardclaim31thatintransitive moueõ is an inner-Latin
development, arising from of
syntacticellipsis reflexivepronouns{sê
mouëre- » [0] mouère,etc.). This claim is notwithoutitsproblems:to

27See
e.g. OLD 3b, Ernout-Meillet s.v., TLL VIII: 1546.9ff.,Hofmann-Szantyr 1965:
290, Kühner-Stegmann 1966: 92 ("Vorklassisch").
28
SimilarlyBammesberger1986 on causative *louk-é¡o/e- and stative*luk-eh¡-con-
flatedin Lat. lücere 'shine' (trans,and intrans.).For the intransitive presentparticiple:
cf. Watkins1969: 144 on inherited"mediopassiveBedeutungdes Partizipsauf -nt-"in
Latin(ëuidëns,uehëns,etc.).
29pj£ *m,-_ > Lat m_regularly.
30For the intransitive
meaningof the originalverb,cf. Toch. B miwam'trembles',
Hitt.mühhi/mauszi 'falls'. (Despite LIV, it is farless certainthatHitt.mummiya- 'scat-
ter,fall apart' belongs withthe materialassigned to ^mjeuh/-;and even if it does, its
formation remainsunclear.)
31 1965: 34f.,followingBendz 1943: 35ff.
E.g. Hofmann-Szantyr

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
220 Brent
Vine

beginwith,thefewparallelsthatare generallyadduced(e.g. occasional


recipere= se recipere'betakeoneself,go') have an altogether different
textualprofile,as comparedwiththewidespreadand earlyintransitive
use of moueõ. The seminaldiscussionof thispatternby Bendz (who
identifiedthisusage forseven verbs,includingmoueõ,in thewritings
of Caelius Aurelianus[5thc. AD]) makesclear thatthisis essentiallya
matterof "volks- und fachsprachlichen Gebrauch" (1943: 35),32a
characterization that would not be easy to defend in the case of
intransitivemoueõin Archaicand Classical Latin.Nevertheless, even if
one wishes to accept this theory,it would actuallytendto supporta
"conflation"approach(as above), ratherthanweakenit. A patternin-
volvingsuppressionof reflexivepronounswithoriginaltransitive verbs
(especially verbs of cf.
motion, [se] recipere) could help explain the
eventualabsorptionof boththeform(synchronie ö-grade,continuing an
older e-grade) and functionof the old intransitiveverb continued
indirectlyin moueõ,accordingto thesuggestionoutlinedabove.
Even if one prefersto insistthatthe only inheritedpresentwas a
causative^miouhj-éio/e- (> Lat. ^maueõ by Thurneysen-Havet's Law):
theperfectof thiscausativecould have been builtfroma transitive s-
aorist*mieuh]-s-withe-grade;compareluxï to lüceö (n. 28 above), as
well as Hittitepresentformslike 2 sg. mausta,perhapsultimately rela-
ted to 5-aoristformsforthisroot(cf. LIV).33In Latin,however,where
such an s-aorist*mieuhrs-would developto ^meua-s-,all vowel-final
5-aoristswere automaticallyreplaced by w-perfects, a process that
would have resultedin *meua-uai> mõul (again, beside supinemõtus
withold e-grade).Subsequentanalogicaladjustments could have taken
place in different
ways,depending on thechronology. Thus, at thestage
près. *maueio/e- : perf. *meua-uai (: supine ^meua-tu-), the present
could have been remadeas *meueio/e-, whencemoueõ. Or, at a later
stage: the resultingpattern*maueõ/mõui(and supine mõtus beside
verbaladj. *mautus< ^mouhj-eto-34) wouldbe highlyaberrant, as com-
pared both withThurneysen-Havet forms like caueõ/cãul (cautus) with
persistent a-vocalism and other old causatives (moneo/monuï/monitus

i¿ Cf. his
judgment about recipere (1943: 37): Der intransitiveGebrauch von
reciperescheintgeradein derMilitärsprache besondersbeliebtgewesenzu sein".
33The
comparisonbetweentheGreeks-aoristáusúoaadai '(ex)change,surpass' and
Lat. moueõ (oftenmootedin the etymologicalliterature, cf. n. 24 above) may well be
specious, if Wächter's attractiveanalysis of áusúoaaSai is correct(cf. ratherajieißco
'(ex)change', replacing*àu£iF(o;2001: 119ff.,cf. S. Minon,CEG 8 [2003] s.v. ausißco).
34This verbal
adjectivemightalso have been in competitionwiththeoriginal*mjuhj-
tó- (> Lat. ~'mütus),which survivesindirectlyin Hitt.mudai- 'remove' (cf. Melchert
1994: 60, withreferences).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' inLatinandItalic
On 'Thurneysen-Havet's 221

etc.) withpersistent o-vocalism;sucha pattern wouldbe ripeforadjust-


mentto moueõ/mõui (mõtus),perhapswithpressurefromthephoneti-
cally similarcausativefoueõ/fõuí(fõtus)(cf. above on thispoint)and,
forthatmatter,the iterativeuoueõ/uõui(uõtus) (neitherof whichwas
ever susceptibleto Thurneysen-Havet's Law, giventheiroriginalroot-
final *-g"' cf. I.2.). Somewhatsimilarly,Rasmussen(1993: 200) en-
visagesformoueõ"reintroduction of thevocalism-o- in whatwas still
synchronically a causative".
In conclusion:besides analogical influencefromfoueõ and purely
phoneticfactors(Meiser's explanations(i) and (ii) above), thereare
viable sourcesforthesynchronie o-vocalismof moueõbased on one or
moreinherited ^-grades,muchas withnouãcula and cloaca.
2.2.4. Similarto moueõis an alleged"'toueo,theexistenceof whichis
sometimesconjectured(thus IEW 1080, Walde-Hofmanns.v. tõtus,
Leumann1977: 133) on thebasis of thefollowingforms:
(i) tõmentum 'stuffing (forcushionsetc.)' [Varro+],formally parallel
to momentum[Ter.+] (cf. moueõ) and fõmentum'(warm) poultice;
alleviation'[Cic.+] {cí. foueõ), and comparableto other-mentum forms
builtto causativestems(moneõ/monumentum etc.,Leumannloe. cit.);
(ii) theadjectivetõtus'all', takenin thisviewto be theoriginalverbal
adjectiveto ^toueõ.
DespiteWalde-Hofmann, theetymology of tõtusimpliedby (ii) is far
fromsecure,and this formis best leftout of accounthere. But even
tõmentum lends itselfto different interpretations.Thus Ernout-Meillet
prefer*ton(d)-s-mentum (to tondeõ 'shear,clip'),35and Perrot(1961:
168 with n. 6) declares thatthe word "n'a pas d'étymologiesûre",
addingevidencesuggestiveof a borrowing fromGaulish.Nevertheless,
thematerialin (i) is somewhatsuggestiveof a causative'toueo 'stuff,
whichdeservesto be evaluatedbrieflyas a counterexample to theex-
of a
pectation Thurneysen-Havet resultcomparable to caueõ.
In short,some of the same argumentsapplied to moueõ are equally
valid for^toueõ,mutatismutandis.The etymologicalbasis of theinter-
pretationin question (thus IEW loc. cit., Walde-Hofmanns.v.
tõmentum)is the root *teuh2-'swell, become strong'of Ved. tavïti
'grows strong'(LIV 639f.,withoutmentionof ^toueõ). For thisroot,
M. Kümmel (LIV) plausiblyreconstructs an athematicroot present
*téuh2-ti/*tuh2-énti (cf. tavïti),withpossible traces of a reduplicated
present(in theVed. aor. inj. 2 sg. tutos,3 sg. tütot,see Kümmel2000:
220f.) and a nasal present(in OP tunuvant-'mighty').Thus, as with

35Cf.the inCGL 11.595.36:


entry 'genusherbaequaeproplumisinlectum
mittitur'.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
222 Brent
vine

moueõ,a corresponding activerootaorist*téu'i2-t 'was swollen/strong'


(> 'was filled/stuffed')wouldhave been regularly replaced,in Latin,by
a «-perfect*téua-uai> *tóua-uaP6> ^tõui. Unlikemoueõ,however,an
inherited causative *touh2-éio/e- 'strengthen, make swollen' (> 'stuff)
- if it everexisted- wouldnotnecessarilyhave developedto an ¿F-verb
"'taueo(like caueõ). Given theLatin ã-verbssuspectedby some of be-
ing originalcausativesto roots withthis structure (i.e. CeCh2-),37 we
might ratherexpectthe result to have been a ^touãio/e- > ^tauãio/e-(by
Thurneysen-Havet's Law) > pianare, with perfectumHeua-uai (cf.
*dema-uai,*suena-uai,etc.) > *toua-uai > 'toul. Especially in that
event,theresulting^tauãre/^tõui(transitive) beside intrans, perf.'toul
(as above) would have been intolerable; and given the formaland
semanticparallelof mõul(bothtrans,and intrans.),38 an evidentsolution
would be the creationof a ^toueõ, whence tõmentum. Nor is a "'toueo
actuallyneededto accountforLat. tõmentum: a perfect"'toui'I am/have
become stuffed'(inheritedfroma PIE rootaoristor perfect,the latter
being the otherreconstructable formfor this root: cf. LIV on Ved.
tütava,YAv. tütauuä)mighthave provideda sufficient basis (by ana-
logyto mõui/mõmentum, fõui/fõmentum) forthecreationof a tõmentum.
In sum,a ^toueõ - which,afterall, is entirelyunattested - need never
have existedat all, and need notin anyeventreflectan originalforma-
tionthatwouldserveas a counterexample to Thurneysen-Havet's Law.
2.3. As discussedin moredetailbelow (4.1., 4.3.), themostwidely-
accepted versionof Thurneysen-Havet's Law restrictsthe process to
*/ou/precedingan accentedvowel, in termsof theClassical Latin ac-
cent (thus cauëre, cauëbam, cauëbo etc. would be phonologically
regular,withcáueõ, cáuet,cáuè etc. theresultof analogy).For such a
theory,root-accented formswithö-vocalism- such as ouis 'sheep' -
are unproblematic. But theconsiderations bearingon thechronologyof
Thurneysen-Havet's Law (1.2.) already make clear thata formulation
based on the Classical Latin accentcannotbe correct.It now appears,
then,thatunshifted formslike ouis (and likewiseformsof thewordfor
'cow': gen. bouis,ace. bouem,etc.) are exceedinglydifficult to account

36NotsubjecttoThurneysen-Havet's
Law,since-ou-hereresults from -eu-(I.2.).
37 Thus domare(domuî,domitus)'tame',uetãre,OLat. uotãre(uetuï,uetitus)
cf.sonarebesidesonere(sonuï,sonitus)'sound',tonarebesidetonere{tonuï,
'forbid',
tonitus) on themorphological
'thunder'; background ofsuchforms see Schrijver 1991:
391ff..Rixl999:517ff.
38Withadditional
formal perhapsappliedbyfoueõ/fõui,uoueõ/uõui.
pressure

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 223

for,just as theywereforearlierpre-accentual theories,beginningwith


theoriginalstudiesof Thurneysen and Havetthemselves.39
2.3.1. For bouis/bouem, a popularapproach(beginningwithThurn-
eysen)appeals to the dialectalstatusof bõs; butif Thurneysen-Havet's
Law is as old as it seemsto be, thiswouldnotbe helpful.Still,thereis
the possibility,alreadymentioned(2.2.3., ad ink.), of a phonological
explanation(and a different phonologicalsolutionwill be proposed
below). It seems possible,moreover, thattheo-vocalismof nom.sg. bõs
(whether or not thatform directlycontinues PIE *guous,on whichsee
Sihler 1995: 335f.) mighthave exertedsufficient pressureto maintain
oblique forms like bouis (vs. *bauis). If, moreover,oblique formsof
thiswordare to be reconstructed withe-grade(cf. Sihler 1995: 334f.),
such forms(dat. *guéu-eietc.) would not have been susceptibleto
Thurneysen-Havet's Law in anycase.
2.3.2. Much more difficultis ouis. The old claim thatthe word is
borrowedfromGreek or Sabellian (thus Havet, Horton-Smith, and
others)is no longeracceptable.Anotherhoaryexplanationassertsthat
an applicationof Thurneysen-Havet's Law would produce^auis, gene-
ratinga case of "homonymie fâcheuse"withauis 'bird'; but Schrijver
properlyobservesthat"if this was the only reason,restoration of ou
seemsunlikely".Schrijver' s own approachto ouis is as follows.Opera-
tingwitha reconstruction ^h3éu-i-(wheretheinitial*h3e-would seem
to predictold ö-vocalismfor Italic), he suggeststhat the sequence
*/h3eu-V-/ (withoriginalold *-e-) mightstillhave been phonologically
distinctfromthe ordinary*/o/of the/ou/diphthongat some stage of
Italic, and thuswould not have been affectedby Thurneysen-Havet's
Law. Partof thebasis forthissuggestionis theLatin versionof Rix's
Law,40showingdistinctivecolorationforthe laryngealsat a stage not
long beforeItalic, as well as a claim by Lubotsky(1990) concerning
*h3e-as phoneticallydistinctfrom*-o- and *//o-in the prehistory of
Sanskrit,41 as a way of accountingforthefailureof Brugmann'sLaw to
applyto Ved. ávi- 'sheep'.42

39 See in detail
Schrijver1991: 436 (general),447 (bouis etc.), 449f. (ouis), 453f.
(conclusions).
40See
Schrijver1991: 56ff.,Meiser 1998: 106.
41In essence
(cf. Lubotsky1990: 135), applyingto Sanskrita versionof Kortlandt's
theory(1980: 127f.)thatArmeniandistinguishesoriginal*h¡o- from*h3e-.This beha-
vior,however,is farfromclearlydemonstrated; see Mayrhofer1986: 141, Olsen 1999:
18n31.47n95.
42
Oddly, even thoughSchrijver(followingLubotsky1990: 133f.) operateswith a
preform*g"h3-eu-for 'cow', he does not avail himself,in his discussionof thatword

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
224 BrentVine

There are, however,a numberof difficulties withthis account.To


beginwith,it seemsabundantly clear thatthecoloringof *h3e-to *h3o-
musthave takenplace long beforeItalic (Meier-Brugger2003: 112), in
which case the distinctiveness of the laryngealsthemselvesat some
stage of Italic (in certainotherpositions)is beside the point;see the
similarargumentation by Rasmussen(1993: 200) againstSchrijver's
conception ("theoretically possible"but"hardlycredible").Nor,as Lu-
botsky himself admits (1990: 130, 134), does the non-application of
Brugmann's Law in Ved. ávi- require a claim aboutinitial*h3e-having
been treateddifferently fromold */Ho/(see morerecentlyKim 2000:
39n4, with references).Finally, the word for 'sheep' is now better
reconstructed as *h2óu-i-,in view of Lycian (ace. sg.) xawa (vs. *h3->
0 in Lycian).43Given,then,theinescapableold ö-gradeof sucha form,
Lat. ouis emergesas the single most intractableexceptionto Thurn-
eysen-Havet'sLaw (cf. Meiser [1998: 85]: "Schwierigbleibtovis.").
The solutionto thisproblem,as I suggestin thenextsection,calls fora
reformulation of Thurneysen-Havet's Law.

3. An accentual theoryof Thurneysen-Havet'sLaw (I):


preliminaryremarks
3.1. Let us firstrecapitulatethe chronologicalconceptionof Thurn-
eysen-Havet'sLaw developed in section 1., as already surmisedby
Schrijverand Meiser,and supportedby a probableinstanceof thepro-
cess in Umbrian:simplyput,Thurneysen-Havet's Law is amongtheol-
destinnovationsassignableto Proto-Italie.Indeed,if *eu > *ou (which
Thurneysen-Havet's Law must precede) is taken to be Proto-Italie,
Thurneysen-Havet's Law may be the earliestdatable developmentof
Proto-Italie.On the basis of this chronology,then,it is possible to
suggestthe followingface-valueinterpretation of formslike caueõ on
the one hand beside formslike ouis on the other:as a firstapproxi-
mation,we may hypothesizethatThurneysen-Havet's Law was condi-
tionedbywhatremainedofthePIE mobileaccent(at a veryearlystage
of Italic),and affectedonly*ou sequencesin unaccentedsyllables.The
followingsub-sectionsexplore,in a preliminary way,theconsequences
of this suggestionfor some of the prominentformsand categories
mentioned so far.

(1991: 447), of an argumentbased on an alleged differencebetween/h3eu/ and /ou/,


comparableto his suggestionfor'sheep'.
43Kimball 1987, Melchert1994: 72, Kim 2000: 38n2, Yoshida 2004: 195.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 225

3.2. For thecanoniccases of thetypecaueõ andfaueõ, theprocessas


just describedwould be regularforthepresentsystemof originalitera-
tive-causatives, undertheassumptionthatthePIE accentuation of forms
like *kouhréio/e-was preservedintact.Similarly,it is reasonableto
assumethatverbaladjectiveslike Pr.-Ital.*kou-eto-, even thoughana-
logically created within would
Italic,44 have had oxytoneaccent (thus
*kou-etó- > Lat. cautusregularly),like theinheritedverbaladjectivesin
*-tó-on whichtheyare ultimatelybased. A root-accented perfectum
*kóua-uai,of course,would not have undergoneThurneysen-Havet's
Law in thisformulation; but thereis littledifficultyin assumingthat
such a formwould have been remodeledto *káua-uai(> Lat. cauï) on
thebasis of the¿z-vocalism of theinfectum and theverbaladjective.45
3.3. Essentiallythe same profileas caueõ appears in fauëre (fauï,
fautus)< *g"hou-éio/e- (followingSchrijver[1991: 442] fortheroot,cf.
OCS govëti 'worship',Arm.govern'praise').46The surprising inscrip-
tionalformFOVE = 2 sg. imperative fauë (CIL I2 573; on a Praenestine
scraper,late 2nd/early1st c. BC) is traditionally takento supportthe
accentualversionof Thurneysen-Havet's Law based on the Classical
Latin accent(2.3.). But the currentproposalrequiresthatthe formbe
explained secondarily,for which several optionsare available. As a
non-urbaninscriptionalhapax froma period that predatesClassical
orthographic norms,the formmay simplybe untrustworthy: Schrijver
(loc. cit.) suggeststhepossibilityof an error,or a local developmentof
olderfauë. As an "error",forexample: one may imaginea contami-
nationwithfouëre 'warm' in its sense 'cherish,caress, foster'(well-
attestedfromthe timeof Cicero).47As a local development:one can
suggestmore specificallythat the formcould be aligned with the
"woc/wos-Gesetz" (Leumann1977: 50), accordingto whichOLat. uaC-
> uoC- beforeaccentedvowel (OLat. uociuos beside uacuus 'empty',

44On the
process,see Jasanoff 2003: 103n30.
45
Similarly,for perf.cauï, Meiser (2003: 233f.). For the sake of completeness,I
mentionheretheonlyseriousalternative to Thurneysen-Havet's Law, namelythetheory
of Szemerényi(1951), who posits a rule *ow> *ãu (as in perf.cauï < *kouï< *kóua-
uaj, withcaueõ analogicalto cauï). Even if such a ruleoperatedin some forms(though
perhapsnot in initialsyllables,cf. octãuus '8th'), thereare manyweightyobjectionsto
thistheory;detailedargumentation contrain Schrijver1991: 437f.,Meiser2003: 233f.
46 LIV (s.v. *aeu- 'laufen,eilen'); butthedifferencein thereconstruction
Differently
of theinitialconsonantis immaterial forour purposes.
47This seems
preferableto the attempt(e.g. Warmington1940: 202nl) to take fove
as theimperativeoffouërein thissense,withelided me and withthefollowingwoman's
namein -Aitakenas a possessivegenitiveinsteadof dative("Caress [me]. [Property]of
Lucia Cornelia...").

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
226 Vine
Brent

OLat. uocãre [cf.Pit. Cas. 527 uocent]foruacãre 'be empty').48 Under


unknownconditions(e.g. dialectally,or with the additional labial
environment providedby a followingglide), therule mighthave been
extendedto initialfa(u)- at a timewhen/- was stillbilabial,49 yielding
fonere fororiginalfauëre. But just as the uociuos-Gesetz, which in its
clearest instancesproduced uncomfortable homophonic clashes (cf.
uocãre 'call' and uocãre 'be empty',hence the Plautine pun cited
above), was givenup not long afterit began to operate,ã fauëre with
the meaning of fauëre would not have survivedlong (cf. fouëre
'warm'). If,finally,theformcontinuessomethingold, it could in prin-
ciple reflectan originalrootthematicpresent*g"héu-o/e- beside theo-
in The and
gradeiterative-causativefauëre.50 phonology orthography of
such formsin Latin pose complexproblems(see Leumann 1977: 135);
if the formhad survivedintoClassical Latin,it mightbe expectedto
appear as Afuere.But OLat. inscriptionalpracticeof about the same
periodattestsspellingsexactlycomparableto FOVE, e.g. IOVENT = Cl.
iuuent(CIL I2 364; "Faliscan cooks" dedication,no earlierthan 150
BC), CONFLOVONT= Cl. confluunt(CIL I2 584.23; SententiaMinu-
ciorum,117BC).51
3.4. Anotherold iterative-causative of the type seen in caueõ and
faueõ may appear in aueõ 'desire' (not treatedin Schrijver'smaterial,
but cf. alreadyHorton-Smith 1895: 45 If. forthis analysis).Although
oftenconnectedwiththerootof Ved. ávati 'furthers', Lat. iuuõ 'help'
etc. (*h2euH-or [LIV] ^h}euH-),52 M. Kümmel(LIV) prefersto assign
it to a *h2eu-'genießen', and specificallyto an iterative-causative
*h2ou-é¡o/e-seen also in Ved. (3 sg. imperf.)ãvayat 'ate'. For theLatin

48 There is no (uoC- > uaC-) pro-


good evidence favoringthe reverseformulation
posed by Szemerényi(1989: 81ff.,similarlySihler 1995: 44).
4VOn the in OLat., see e.g. Sturtevant1940: 163, Maniet
probablebilabial value ot 1X1
1975: 27; a moreskepticaljudgmentin Allen 1978: 34, followedby Stuart-Smith 2004:
47. It is clear,at any rate,that<f> could be used to spell a bilabial segment,as in the
OLat./dialectalprepositionaf 'ab' (Vine 1993: 175ff.)and Pael. af- (voiced bilabial
fricative,see Stuart-Smith 2004: 120), and thatLat. lïl was likelyto have had bilabial
variants(thusStuart-Smith).
DVA
nearlyidenticalattestationpatternis toundwiththe otherclass ot Latin e-verbs
(i.e. statives),in thetypeOLat. fulgerevs. laterfulgëre'shine' (Leumann 1977: 544).
51Such OLat. formsthemselvesare not Law, as
problematicforThurneysen-Havet's
theymay go back to e-gradesin some cases, and in othersmay merelyinvolve hyper-
correctspellings(as I hope to discuss elsewhere).Several Lat. fau- and fou- formsof
uncertainetymology(and in some cases unclear meaning) are of no value for the
discussion(despitetheirhavingfigured,sometimesprominently, in earlierworkon the
problem),and need not be consideredhere: see Schrijver'scommentson fauis(s)ae,
Favonius,Fouii/Faui,fauus,fouea. Forfauilla, however,see 5. below.
32See
Schnjver 1991: 47 foretymologicaldiscussionand references.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 227

vocalism,Kümmel (ad loc.) writes:"Mit analog. R(e) oder mit lat.


Übergang*öm> auT In view of theabove discussionof e-gradeitera-
tive-causatives(2.2.3.), the latterinterpretation (which worksequally
well whetherone operateswith^ouH-éio/e-, ^¡ouH-éio/e-, or *h2ou-
éio/e-)is to be preferred.
3.5. Notethatin thisrevisedformulation of Thurneysen-Havet's Law,
nouus itself(cf. 1.2.) is not so clearlyprobative,since a root-accented
Italic *nóuo- (< PIE ^néu-o-) would not have been subject to the
process(even ifThurneysen-Havet's Law had operatedon theoutcome
of PIE *-eu-> Ital. *-ow-),and thus mightbe expectedto remainintact
in any case. But an easy substituteis available, namelyLat. nouitãs,
which,in view of Gk. v8Ótt|t-, is likelyto be an inherited item{*neu<5-
teh2t-)thatwould nothave had rootaccent.53 The same functioncould
be performed as well by thefactitiverenouãre(n. 17 above), cf. further
nouãcula and clouãca (2.2.I., 2.2.2.). A possiblenon-Latinexampleof
this sort would be SPi. súhúh (TE 1) (< abl. sg. *souõd, to *seuó-
'self s, own'; accentuation as in Horn,éóç 'id.').
3.6. To return to Lat. ouis: an important consideration is thattheform
goes back to an aerostatic /-stem,54 therefore with ^h2óu-i-in strong
forms(e.g. nom.sg. *h2óu-i-s > Lat. ouis regularly,withoutapplication
of Thurneysen-Havet's Law), and *h2éu-i- in weak forms(> Lat. ^aui-).
Thus, insofaras a homonymieclash withaui- 'bird' mighthave been
involved,thiscould have been responsibleforthegeneralization of the
strong stem-form oui-, even though Latinmay show a to
tendency gene-
ralize the weak stemin o/e-acrostatic paradigms:thuse.g. pês 'foot'
(*pód-/*péd-), genü 'knee' {*gón-u-/*gén-u-)' in contrast,
however,nox
with
'night'{*nógu-t-/*négu-t-), strong form generalized.
3.7. As for 'cow': ö-vocalism,with no applicationof Thurneysen-
Havet's Law, would again be theregularresultforall formsthateither
retainedor developed root accent in the early stage of Italic under
consideration.Thus ace. sg. (*g"ora - ») *guóu-m, nom. pl. *guóu-es,
gen. pl. *guóu-õmwould all have retainedö-vocalismregularly.55 The
Italic accentuationof gen. sg. (*g"óu-sor *g"éu-s->) *g*ou-osand dat.

53 For the accentuationof such formations, cf. Ved. devátãt- 'divinity',etc. (The
medialvowel weakeningin nouitãs,of course,arisesat a muchlaterperiod.)
34 On the aerostaticinflection,see Kim 2000
(especially on Toch. B ãuw, pl. awi
'ewe', based on the oblique stem *h2éu-i-),withearlierreferences, and Yoshida 2004
(especiallyon Lye. yawa-).
55If Sihler
(1995: 334ff.)is correctin reconstructing
o/e-acrostaticinflectionforthis
word,originale-gradeformsmighthave been leveled out in Greekand Latinin favorof
thestrongstem,as he notes:thusgen. sg. *g"éu-s- » pre-Lat./pre-Gk. etc. But
*gk'ou-os,
cf.2.3.I., ad fin.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
228 BrentVine

sg. (l*guéu-ei- ») *guou-eiis perhapsless clear; butif rootaccentwas


notoriginal,itwouldeasilyhave been acquiredbyanalogyto therestof
theparadigm.

4. An accentual theoryof Thurneysen-Havet's Law (II):


furtherconsiderations
4.1. As alreadynotedbriefly(2.3.), theproposaloutlinedin section3.
does not constitutethe firstaccentual theoryof Thurneysen-Havet's
Law. An earlier version,based on the Classical Latin accent, was
suggestedindependently by Kretschmer and Solmsen,56 and thattheory
remainsthe standardone eitheradvocatedoutrightor given pride of
place in most handbooksof Latin historicalgrammar.57 Accordingto
this version,Thurneysen-Havet's Law operatedonly on pretonicsyl-
lables (i.e. on theloi of sequences/ouV/),in termsof theClassical La-
tinpenultimate accent- hence óuis withno application,as opposed to
pretonicauillus 'agnus recentispartus'(P.F. 13.14L [14M]), andfóueõ
'warm' withno application,as opposed tofauïlla 'ashes, embers'.But
theprobableage of Thurneysen-Havet's Law (as discussedabove) is in-
compatiblewithsuch a formulation: the process is farolder thanany
period close to the developmentof the penultimatestresssystemof
Classical Latin. Moreover, some of the key examples must be
interpreteddifferently, as already seen for foueõ (I.2.): the non-
of
application Thurneysen-Havet's Law in foueõ (< *dhog"h-éio/e-),
withthesame iterative-causative formation as caueõ andfaueõ, is better
explainedby theassumption thattheprocessdid notaffecttheoutcome
of *-oguhV- > -ouV-' and therefore, if the 0-vocalismoffauïlla arises
fromThurneysen-Havet's Law, the wordwould have nothingat all to
do withfoueõ,as discussedbelow (5.).
There is, however, an attractiveand importantfeature of the
Kretschmer-Solmsen theorythatremainsintactin the above formu-
lation - namely,the phoneticand typologicalnaturalnessof vowel-
qualityneutralizations in unstressedsyllables(see recentlyFlemming
2005, withfurther references).This includes,to be sure,pretonicun-
rounding,but also, in principle,posttonicunrounding, as in Modern
Russian, where such effects(e.g. nom./acc.sg. okno [aknó], gen. pl.

56 Referencesand discussion in
Collinge (n. 5 above) and Schrijver(1991: 436ff.,
45 If.).
57 See 1977: 159ff.,Leumann 1977: 49f.
e.g. Sommer 1914: 109, Sommer-Pfister
(with the accentualconditioningmistakenlyattributed to Thurneysenand Havet). An
exceptionis Sihler(1995: 44), witha veryloose formulationwithoutreference to accent
("öv (of anyorigin)sporadicallybecomesav").

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 229

okon[okan] 'window(s)',etc.) belongto thebroadersyndrome of vow-


el reduction(i.e., neutralization of vowel quality) in unstressedsyl-
lables. Parallels involvingthe glide environment can also be found.
Thus,in earlyOld Irish(8thc.+), an /ou/> /au/processin accentedsyl-
lables shows the same (descriptivelydissimilatory)unrounding.58
Similarly,in BritishCeltic, thereis evidence for a developmentof
PrimitiveBritish*-ow-to Middle Welsh -aw-,apparentlyrestricted to
Late Prim.Br. penultimatesyllablesand possiblyconditionedby the
vowelqualityin theonsetof thefollowingsyllable(see Schrijver1995:
336ff.).59Despite thedifference in the accentualconditions,thesepat-
ternssharean important featurewithThurneysen-Havet's Law, in terms
of theirsyllablestructure. In all of these "/ou/ > /au/" processes,the
glide is followedby a vowel, and thusthe *-o- and the glide are hete-
rosyllabic(thus/o$uV/> /a$uV/,where"$" denotessyllableboundary).
This configuration shouldbe consideredtogetherwithtwo otherfac-
tors,the firstof whicharises fromthe natureof roundingitself.In a
recentanalysisof thetypologyof roundingharmony, Kaun (2004) has
surveyedthe phoneticpropertiesof roundedvowels, summarizing re-
cent studiesthatdescribetheirarticulatory, acoustic, and perceptual
features.Fromall of theseperspectives,it emergesthatthemanifesta-
tionsof roundingare nonuniform withrespectto vowel qualitydistinc-
tions(i.e., in termsof tongueposition).Most important forourpurposes
are correlationswith vowel height,includingthe followingobser-
vations:in termsof articulation, themagnitudeof lip roundingtendsto
be relativelygreaterforhighvowels thanfornonhighvowels; in terms
of vowel acoustics,lip rounding"has more dramaticacoustic conse-
quencesforhighvowelsthanformidvowels" (Stevens1998: 294); and
in terms of perception,nonhighvowels tend to be perceived as
relativelyless roundedthanhighvowels.It is reasonableto assumethat
thesecorrelationscould well have been operativewithrespectto the
PIE and Proto-Italiemid rounded*o as opposed to the glide *w(the
non-syllabic counterpartof highrounded*w).
The furtherconsiderationthat the spread of phoneticfeaturesis
inhibitedby syllableboundarymeansthatanyroundingeffectof lui on
loi in heterosyllabic/o$uW would ordinarilybe weaker than in
tautosyllabic/ou$/. Given, then, the typological facts about the

58
E.g. Oír. naue 'new' (Sg. 5b6.217),vs. Gaul. Neuio-ZNouio-etc. (Thurneysen1946:
4Sf V
59
E.g. MW llawer 'large number,many',vs. MCo. lower,lowr 'many;enough', cf.
Oír. loor, lour 'enough,sufficient'< Prim.Ir. Houero-< *ro-uer-o-(cf. Oír. rofera 'is
sufficient').

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
230 Brent
vine

relativelyweaker roundingof mid vowels as compared with high


vowels,theroundingof loi in heterosyllabic /o$uV/may in some lan-
guagesbe relativelyunstable.Pressuretowardsunrounding ofloi in this
positionmightalso be enhancediftheloi is unaccented,sincerounding
gesturesin shortervowels would be less complete(and unaccented
vowels tendto be shorterthanaccentedones). Thus, forexample,the
roundingin accented/o$uV/(cf. Pr.-Ital.*óui-< PIE ^h2óu-i-'sheep')
is "protected",so to speak, by the accent,as opposed to unaccented
/o$uV/.60 If thecontrastbetweenrelativelyunrounded loi in /o$uV/and
moreroundedloi in /o$uV/becomesphonologized,theresultwouldbe
whatappears,on thesurface,to be a "dissimilatory unrounding" of loi
followedby theglide lui, butwhichmayinsteadhave moreto do with
thephonetictypologyof roundedvowelsand withfeature-spreading be-
haviorwithrespectto syllableboundary.61
These hypothesesare also of interestwhenviewed froma historical
perspective.To considera formlike caueõ: theessentialdifference bet-
ween PIE */kou$hiéie-/ and Pr.-Ital.*/ko$uéie-/is the shiftin syllable
structure fromtautosyllabic/ou$/to heterosyllabic /o$u/,as a conse-
quence of theloss of theprevocaliclaryngeal.One could theorize,then,
thattheThurneysen-Havet loi > IdJ"effect"is in some way relatedto
the dramaticallyincreasedfrequencyof morphemes,in Proto-Italie,
with"unprotected" loi (beforelabial glide) in heterosyllabic position.
Indeed,as we will see, manyThurneysen-Havet formsinvolvejust such
a background,with root-finallaryngeallost beforevowel in Proto-
Italic.62
4.2. The presentmodel, then,representsa "modifiedKretschmer-
Solmsentheory",based not on the Classical Latin penultimate accent,
but on the remainsof the PIE mobile accent at some early stage of
Italic.Note furtherthattheproposedformulation, forreasonsthatwill
become apparent below, operates more broadly with a structural

60
Compare,in English,thecontrastbetweenloi underprimarystressin theadjective
thrówaway(always realized as [o]) and loi preceded or followed by primarystress,
respectively,in Genoa and Halloween,whererealizationwithschwa is normalin casual
speech.
61As forthe
precise mechanismthatmightaccount forthe phonologizationof this
rule (a topic thatcannotbe exploredhere): this would depend on one's views about
historicalphonologicaltheory;severalcurrentapproachesare convenientlysurveyedby
Stuart-Smith (2004: 8ff).
oz I am
extremely gratefulto Adam Albright(MIT) and Bruce Hayes (UCLA) fordis-
cussion and referencesconnectedwiththe phoneticand phonologicaltreatment in this
section.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' inLatinandItalic
On 'Thurneysen-Havet's 23 1

description of therulethatallows fortheloi of anyunaccentedsyllable


to be affected, and notjust loi in pretonicsyllables.
The generalapproachthatunderliesthismodelrestson two mainas-
sumptions, neitherof whichseemsdifficult.
4.2.1. The firstis simplythatsome remnantof theIE systemof mo-
bile accentsurvivedintoan earlystageof Proto-Italie, beforethewell-
known(and later) period of initialstressthatcharacterizesboth pre-
Latinand theprehistory of theSabellianlanguages(as well as Etruscan,
duringpart of its recorded history).If thisassumptionis correct,Italic
would be roughlycomparableto Germanic,and Thurneysen-Havet's
Law would have a statussimilarto Verner'sLaw, withthe exception
that- forevidentreasons,suchas theprevalenceof subsequentsyncope
rules and monophthongizations in a large partof Sabellian, and to a
certainextentalso in Latino-Faliscan- thereare farfewervisibletraces
of Thurneysen-Havet's Law in Italic (and especiallyin Sabellian) than
thereare of Verner'sLaw in Germanic(and, partlyas a consequence,
no systematicexploitationof the process in grammaticalcategories.)
Still,if thisgeneralpictureis correct,Thurneysen-Havet's Law would
providea valuablewindowontothestateof thePIE accentin Italicfor
a numberof individualformsand morphological categories.
Here itcan be mentioned, in passing,thatthesystemof mobileaccent
thatsurvivedinto earlyItalic need not have involvedthe same pitch-
based systemnormallyreconstructed for PIE itself.On typological
grounds,given the natureof thevowel-quality neutralization
rulebeing
proposed, it is rathermore likely that the accentual in
system question
alreadyinvolveda mobilestressaccent,roughlyas in ModernRussian
(as comparedwithProto-Slavic),ratherthana pitchaccent.
4.2.2. The second major assumptionis thattracesof this surviving
mobilesystemmightwell be preservedin Latin.Indeed,morethanone
claim aboutsuchtracesalreadyexists.Perhapsthebestknownof these
assertsthatsome Latin prehistoricsequences of the shape CaRaC go
back to root-stressed sequencesCRHC, as opposed to thedevelopment
of unaccentedCRHC to CRãC.63Thus Lat. palma 'palm,hand' would
resultfroma syncopated*palamã < (root-accented)*plh2-meh2 (cf. Gk.
7tatax!ir|'id.'),64vs. the regularunaccentedtreatment in plãnus 'flat' <
*p¡h2-nó-. Anotherproposalof thissort(Rix 1996: 158n7)concernsthe
treatment of shortvowels in open finalsyllables,as in et 'and' < *éti
63Fordata,earlier anddiscussion,
references, see e.g.Schrijver
1991:193ff.(contra),
Meiser1998:108f.(pro).
64The same forGreekis widelyaccepted(including see e.g.
itself),
pattern 7caÀ,áur|
Rix 1992:73; a dissenting
viewinSihler1995:107f.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
232 BrentVine

(with apocope of the unaccentedfinal vowel, just as in the active


primaryendingsof verbs),as opposed to ablativesingularpede, from
locative singular*ped-i,with the originally-accented final short *-i
preserved on the surface (with altered quality); this final vowel is then
generalized to other categories where the locative singularwould not
have been accented- hence genus 'race, kind' withabl. sg. genere(<
*génhres-ï)forpredicted"f genis,etc. This patternhas been discussed
by Meiser (1998: 74), who observes:"Das Lautgesetzist insofernbe-
merkenswert, weil es als einziges spezifischitalischesauf die Ak-
zentuierung vor der gemeinitalischen Festlegungaufden Initialakzent...
verweist."
Neitherof thesetwo claims is necessarilycorrect(a subjecttoo com-
plex to explore here); but both are defensible,and in some respects
attractive.The currentproposal,then,is thatThurneysen-Havet's Law
may provide another such example pointingto the possibilityof
reconstructing somefeaturesof a Proto-Italie mobileaccent.
4.3. The originalversionof the Kretschmer-Solmsen theoryentailed
anotherseriousproblem,apartfromtheone alreadynotedhavingto do
withrelativechronology.Underthattheory,therewere in fact many
directcounterexamples of bothtypes- thatis, root-accented formswith
tf-vocalismsuspected of derivingfromold ö-grades (as in lauere,
cáuus, etc.),and pretonicformsthatfailedto undergotheprocess(as in
mouêre,fouëre,etc.). Naturally,thisrequiredan extensivesetof claims
aboutanalogicalreshuffling - indeed,sucha vastamountof essentially
arbitraryreshuffling that, as Schrijverhas made particularlyclear
(1991: 437 and passim in his discussion),a theoryoperatingwiththe
Classical Latin accent cannot possibly explain how so many coun-
terexamplesof the typesinvolvedarose in such a shorttime.For the
"modified Kretschmer-Solmsen theory" under considerationhere,
operating at the level of Proto-Italie,some of the very same
counterexamplesinevitably surface. In general,however,analogical
-
reshufflings and, it is hoped, of a less arbitrary nature- can be
countenancedwithrather less difficulty,given the much greatertime
depthinvolved.
4.4. A thoroughdescriptivetreatment of all the formsthatmightbe
relevantforan assessmentof thistheoryis hardlynecessary(even if it
were feasiblewithinthe scope of an article),especiallysince mostof
themhave been discussed in detail by Schrijver.65 It is nevertheless

65 Not all of his


analyses are equally convincing;thus,for formsthatneed not be
assessed in detail here: if caudex 'tree trunk'belongs withthe 'hew' root,Schrijver's

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 233

appropriate to offera preliminary siftingof someof themostsuggestive


materialforthetheory- as well as some problematic data - in addition
to theitemsalreadytreatedabove.
4.4.1. Of the gloss aububulcus 'pastor bouum' (CGL V. 346.39),
Schrijverwrites(1991: 439) thatwhetherit containsau- < *oui- is
extremely uncertain,giventhecorruptnatureof thegloss, thefactthat
the word is given a meaning'cow-herd',and the lack of evidence in
Latin for dvandva compoundsof this type ('sheep-/cow-herd'). The
second point,however,has littlemerit:the meaningprovidedby the
glossatoror adjustedby a subsequentcopyistmay be affectedby the
textualhistoryof theglossedword,whichat somepointin thetradition
may no longer have been understood.The final objection is more
weighty(althoughLindner[2002: 219] seemspreparedto acceptsuch a
compound for Latin, comparingGk. i7r7roßoDKOA,oc for the type66).
Nevertheless,Bährens' attractivecorrectionau[bu]bulcus 'pastor
[b]ou<i>um' (i.e., aubulcus 'pastorouium') providesan elegantsolu-
(see alreadyHorton-Smith
tionto all difficulties 1895: 453, withearlier
references, and especiallySaussure1879: 104). The onlysupplementary
assumptionneeded, under the currenthypothesis,is that when the
originalcompoundwith*oui-was formed(cf. bubulcus'cow-herd'and
subulcus'swine-herd';Lindner2002: 218f.),thefirstmemberof sucha
compound was unaccented. While there can be no independent
confirmation of thisfromLatin,comparativeevidencecertainlyfavors
such a claim,whatevertheprecisebackgroundof the obscureelement
-bulcus: cf. ßo')KOAx)c 'cow-herd',aÍ7tóta)ç'goat-herd'(both paroxy-
tone, insteadof oxytone,by Wheeler's Law), xxpopßoc'swine-herd',
Ved. avi-pãlá- 'shepherd' (also aja-pãlá 'goat-herd',go-pãlá 'cow-
herd'),go-pã- 'protecting' (< *'cow-herding').
In the securely-attested typeof Latin dvandva,the compoundmem-
bersprecedea derivationalsuffix,e.g. strüfer(c)tarius 'priestin charge
of the strues and fer(c)tum[sacred cakes]' (Leumann 1977: 403,
Lindner 2002: 24, 91). If Italic followed the normal suffix-accent
patternof dvandva-accentuation in Sanskrit(cf. Ved. ajãváyah 'goats
and sheep'), themedial-oui-(forexpected-'-aui-)in thefamousthree-

*kouh2dh-ek- (> pre-Lat. *kauadek-by Thurneysen-Havet'sLaw, 1991: 287) would


violateSaussure's (laryngeal-loss)Law; therootwas of the"long-diphthong" type(LIV
2. *keh2u-),so thatan original *keh2u-d-is equally possible (Vine 1982: 75f.). Or:
despiteSchrijver(1991 : 446), thereis littlereasonto suspectoriginalo-gradesin pauire
'beat' and pauëre 'be struckwithfear'(see Hackstein1993: 154, 161); etc.
66 On the of ircTtopouKOAxx;
interpretation [Soph., Eur.] (probably 'Pferdehüter'):
Panagl 1999:442.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
234 BrentVine

partdvandvasuouitaurïlia'sacrificeof a pig,sheep,and bull' wouldbe


an analogicalrestoration aftersimplexouis - and a trivial(perhapsvir-
tuallyautomatic)one, thestatusof sheep in thissacralterm(and in the
ceremoniesto whichitbelongs)beingquiteclear.Adjectivesof thesort
ouillus (4.3.), ouînus, ouîlis, involvingproductiveformations, could
likewise show analogical restorations afterouis, even thoughoriginal
oxytoneaffixes(such as adjectival *-nó-in ouinus) mightwell have
beenpreservedas suchat thestageof Proto-Italie in question.
A somewhatmorecomplexpictureappearsin õpiliõ 'shepherd'[Pit.,
Cato+],67beside the variantüpiliö (e.g. V. E. 10.19); the latteris now
generallyassumed to be the "urban"variant,and the former"rustic"
(see Leumann1977: 134, followingErnout1909: 46, 209, cf. 1905/6:
329f.), both somehow involvingoui- as the firstmemberof an old
compound.If thisis so, however,theotherwisedifficult alternation õ-
/ü- could be interpreted naturallyby assumingthatõpiliõ reflectsa
dialectal developmentof an original*aupiliõ (< *aui-p°),6S while the
"urban"variantüpiliö is the regularresultof a versionwithrestored
firstmember*oui-, as in ouillus etc. above.69In orderto justifythe
applicationof Thurneysen-Havet's Law in *aupiliõ,we mustconsider
the formation of the compound,whichthe communisopinio takes to
containa second memberrelatedto Lat. pellõ 'push, strike',hence
originally'Schaftreiber'(see Walde-Hofmanns.v. and Gaide 1988:
21Of.,withreferences). If thatis correct,thethematic(post-syncope and
post-vowel-weakening) *aupilo- on which the Lat. mase, -iõ formation
could be based (see Leumann 1977: 365) could itselfgo back to a
*oui-pol-o-(< ^h2oui-polh2-ó-) 'sheep-driving' (fortheaccent,cf.again
avi-pãlá-,10ixpopßoc etc. above). Alternatively, given formslike Ved.
go-pã- (above), Arm. hoviw <
'shepherd' *h2oui-péh2-ylx one could also
imagine a source of the type*h2oui-ph2-ló- 'sheep-protecting' (see next
on
paragraph oxytone*-/d-).72 In eithercase, the unaccented firstmem-
a of
ber would condition regularapplication Thurneysen-Havet's Law.
It is convenientto note, at this point,that a Venetic example the
of
*h2oui-polh2-ó- type (except withthe Thurneysen-Havet formas the
in
secondmember)mayappear thepersonal name ho.s.tihauo.s.,if one
67Also Opilius/Opiliain Praenestine
graveinscriptions.
08See
e.g. Leumann1977: 72f. on Lat. au/õ.
69
Essentiallythesameconceptionalreadyin Juret1913: 240.
70 With no actual connection,
despite appearances,to the *-pol-o- of Latin; see
EWAiaII.124s.v./?fl/a/.
71On this
compoundtype:Pinault1993: 214.
72For
õpiliõ in termsof second memberbased on *peh2-'protect',see e.g. Osthoff
1894: 282 (thoughwithfalsecomparisonto Ved. -pãlá-, cf. n. 70 above).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
s Law' in Latinand Italic
On 'Thurneysen-Havet' 235

accepts Watkins' attractiveetymologicalinterpretation *ghosti-ghouó-


'who honorsguests'(1995: 246 withn. 14; cf.Prosdocimi1978: 265 on
*-c?w-> -au- in Venetic,thoughtheexamplediscussedthere[theonym
Plavis] is considerablyless certain). The importanceof a Venetic
exampleforthechronologyof Thurneysen-Havet' s Law is evident(cf.
1.2. ad fin.).
As foravillus (4.1.): Schrijver(1991: 438f.) has added additionalar-
gumentsto the earlierview (acknowledgedalreadyby Horton-Smith
1895: 453) thatthewordmightin factcontinue*ag*n-elo-(thusbased
noton thewordfor'sheep' butratheron thewordfor'lamb', as might
be expectedin view of its meaningand gender).If thisis correct,the
alternationbetweenouis and auillus, one of the cornerstonesof the
Kretschmer-Solmsen theory,becomes entirelyspecious. If, however,
one assumes thatthe originaldiminutiveformationof the word for
'sheep' involvedan oxytone*-/d-thatsurvivedas such intoItalic,then
the resulting*aui-ló- (< Transponat*h2oui-ló-)could have servedas
thebasis fora doublediminutive (formedwith*-elo-,as usual in Latin)
that would yield auillus regularly.To be sure, the PIE status of
diminutive /-formations is a matterof debate; see recentlySchuhmann
2003, who arguesagainsta PIE diminutiveformant*-/<?- as such. But
even if such diminutives arose independently in Italic (as elsewhere),
theunmarkedassumptionis thattheirsourcewould be thefamiliarad-
jectival *-/d-,which (like *-rd-,*-m5-,and othersuch thematicsuf-
fixes)was normallyoxytone:cf.RV vrsalá-(X.34.1 Id) 'poorfellow' (:
vrsan-'man') (Wackernagel-Debrunner 862f.,withmanymoreexamp-
les, mostlyoxytone).
4.4.2. Lat. cauus 'hollow' [Cato+] has always been taken to be a
Thurneysen-Havet' s Law form,withan originalo-grade^ítouH-o-, to
*£euH-'swell', withderivedwordsfor 'vault,hole', etc.), comparable
to the Greek formsattestedin the glosses KÓor KoiÀxòuaia,kóov xà
Xáouaia ttjç yfjç(Hsch.). Lat. cauus, however,was a primecounter-
example to the Kretschmer-Solmsen theory(4.3.), and if the Greek
materialaccuratelyindicatestheoriginalpositionof theaccent,73 cauus
remains,at firstsight,a primecounterexample forthepresenttheoryas
well.74But it may be significant thattheseGreekformshave substan-

73 That is, insofaras the notationof accent in the MSS of


Hesychius(notoriously
insecurein thisregard)can be trusted.
74
Schrijver,in his analysisof ouis (1991: 450), brieflyentertainedthepossibilitythat
like caueõ could in theorybe accounted
theloi of ouis and the/a/of iterative-causatives
foron thebasis of thePIE accent;buthe rejectedthisapproachpreciselybecause of its

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
236 Brent
Vine

tivalmeanings:thus,if theaccentis takenat face value, root-accented


KÓoican simplyreflectthe regularGreek accent-retraction thatoften
accompaniessubstantivizations derivedfromoriginaloxytoneadjec-
tives (in the presentinstance*ko(|:)óç).75In thatcase, the adjective
itselfcould well have been a "tojióç form"rfouH-ó-'swelling',theIta-
lic reflexof which(accordingto thepresenttheory)wouldtherefore be
subject to Thurneysen-Havet's Law and develop regularly as Lat.
cauus. Or, to view the mattersomewhatdifferently (but withthe same
result):kóoi could show the tójioç form beside the adjectivaltojióç
formon whichLat. cauus itselfis based. It would in any case be ex-
tremelyinteresting if Latin could be shown to be a language that
enables us to distinguish betweenxófioçand tojióç forms- at least in
some cases - on the basis of the application(or not) of Thurneysen-
Havet's Law, a possibilityto be consideredfurther in the next two
subsections.76
Even if one wished to insist on an originalroot-accentedo-grade
^óuH-o-, analogical sources for the a-vocalism of cauus are con-
ceivable, such as remodelingafterthe factitive*kauãió/é-> cauãre
[Varro+].As forcauerna 'hollow,grotto'[Varro,Cic.+]: giventhepro-
bable Etruscan source and relativelylate date of the Lat. -erna
formations (see e.g. Leumann1977: 322), the wordcould simplyhave
been formedon thebasis of thealready-existing *kauo-thatappearsin
cauus.
A special problemconcernsa series of surprising o-gradeformsin
Ibero-Romance(e.g. Port, covo 'concave, hollow, deep', cova 'pit,
cavity,hole', Span, cueva 'cave', Cat. cova 'id.'; Meyer-Lübke1935:
173), which cannotderive directlyfromLat. cauus. Under previous
conceptionsof Thurneysen-Havet's Law, it has been difficultto ac-
count forthese formsin any reasonableway; especiallyunsatisfying
was the view of these formsas outlyingsurvivorsthat somehow
escaped theoperationof thesoundlaw (see Schrijver'scritiqueof this
position[1991: 440], buildingon Solmsen). Underthepresenttheory,
however,theycould readilybe explainedas relicsof root-accented by-
forms,such as the tójuoçform^IcóuH-o-(cf. kóoi), beside the more
widespreadtojióç formthatunderliescauus. Alternatively, in view of

apparentinabilityto explain cauus, on the assumptionthatkóoi shows the original


accent.
75Several
examplesof thisprocessappearedin n. 16 above.
76For recentdiscussionsof
xóuoç and touóç forms,see especiallyKrasukhin2004
and García Ramón 2004: 145ff.(the latterincludingLatin material:accentuallyam-
biguoussulcus 'furrow'beside Gk. óàxóç 'id.').

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' inLatinandItalic
On 'Thurneysen-Havet's 237

thesubstantival meaningand formation of thebetter-attested feminines


(Port,cova, Span, cueva, etc.), theseformscould pointto an original
substantive*ícóuH-eh2, derivedby accent-retraction (cf. above forthis
derivationalpatternin Greek,and 4.2.2. forthe possible retractedac-
cent in Lat. palma). In thisevent,adjectivalograde formslike Port.
cóvo would show trivialanalogical remodelingbased on this sub-
stantive.
4.4.3. Anotherapparentcounterexampleis rauus 'hoarse' (P.F.
355.3L [283M]), which may well belong withthe *h3reuH-of Ved.
ruváti'roars' etc. (Schrijver1991: 258, cf. LIV "h3reuH-).An evident
solution,in view of theabove discussion,wouldbeginfroman original
xofióçform^rouH-o- *'roaring',as would be semanticallyattractive
in any case. The parallelabstractrauis [Plt.+] 'hoarseness'could not
derivefroman old root-accented *h3róuH-i- directly;butif sucha form
was inherited(> *row/-), it could have been remodeledafterthematic
*rauo-.Alternatively, and perhapsmostlikely,theabstractcould have
been formedwithinItalicbeside *rauo-(thusNussbaum2004: §4.2.2),
given the profusionof such /-abstracts in Latin (cf. Nussbaum 1999,
2003). The derived adjective raucus [Plt.+] 'hoarse' (< ^rauiko-,
Leumann1977: 340f.)wouldthenbe based on *raui-.
4AA. A similarpatternmayappearin saucius (1.2.). As I have argued
elsewhere (2004: 624), the most reasonable morphologicalbasis
involvesan original*ksouo-'scraping',substantivized (- » 'a scrape,
wound') with a velar formant *-Vk- (as in othersuch terms:rãmex
'blood vessel,hernia',uarix 'varicosevein'famex 'bruise,contusion',
etc.); theresulting*(k)sau-Vk-or (withsyncope)*(k)sau-k-wouldthen
serveas thebasis fora derivedpossessiveadjective*sau-k-io-'having
a wound,wounded'.It can now be suggestedmorespecificallythatthe
originalo-gradewas a xofióçform*ksou-ó-(> *ksauó-regularly);for
thesemantics, comparetojióç itself.77
4.4.5. Not everywordwithaw-diphthong arisesfroman applicationof
Thurneysen-Havet's Law; but this kind of source is indeed generally
assumedfora seriesof difficult Latin forms,mostof whichlack clear
etymologies,characterizedby aw-diphthong in the root and dental
'
suffix, e.g. laud- Upraise',
fraud-(cf. U.fros-) 'deception'(see Schrijver
1991: 444 on thesetwo). In theory,some suchitemscan involvesecon-
darydentalstemsbased on thematicstemsthatwereoriginallyoxytone,

77 Even if one chooses to


operatewithan accentedö-gradesource(such as
therootvocalismcouldhavebeenremodeled
*ksóu-Vk-), aftertheItalic*ksauéio/e-
discussed
attheoutset(1.2.).

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
238 Brent
Vine

and thus susceptibleto Thurneysen-Havet's Law. Thus for dautia


'hospitality etymologicalanalysisvia the *deuH-or
gifts',an attractive
*deh3-u-of Ved. dúvas- 'reward' etc., includingan appeal to Thurn-
eysen-Havet'sLaw forthe rootvocalism,has recentlybeen suggested
by Driessen(2003: 354f.). One could,forexample,beginwitha TOjaóç
form*douH-ó-'bestowing'> *dauo-,whichcould thenhave servedas
thebasis fora i-stemnoun *dau-et-'bestowal',concretizedas *dau-et-
iio- (whencepl. dautid).1*
Anotherapproachmightoperatedirectlywithoriginalathematicpara-
digms thatincludedsuffix-accented forms,such as proterokinetic or
hysterokinetic i-stems or ¿/-stems- cf. Rieken's analysisof forms like
Lat. teges 'covering'in termsof a proterokinetic (1999:
*tég-õt-/*tg-ét-
92), or theisolatedhysterokinetic79 d-stemof Ur. *sarád- 'cold season'.
In such cases, however,ö-vocalismin the root is difficult to account
for.
For some forms,as fora numberof othersabove, analogical remo-
delingmay be a reasonableoption.Thus, givenan old alternation in-
volvinga noun Hóu-ed-'praise' (cf. OHG Hod 'song' < Heu-t-)beside
its Thurneysen-Havet denominativeHauedãió/é-,the noun could per-
haps have been remodeled to Hau-ed-afterthedenominative.80 Still,the
mostnaturalbasis foran ed-stemHou-ed-(or ratherHau-ed-) is a the-
maticHou-o-,whichcould easily have been an oxytoneHou-ó-(> Ital.
Hauo- [- > Hau-ed-]regularly).
While such formscertainlypose difficulties, in partbecause of their
uncertainetymologicaland morphologicalbackgrounds,none consti-
tutesa directcounterexample. Note,finally,thatiffraud-is alreadyat-
testedwithfl-vocalismin line A3 (]OS : FRA[) of the "Corcolle Altar
Fragments"(CIL I2 2833a, ca. 500 BC), and if the a-vocalismresults
fromThurneysen-Havet's Law (cf. Vine 1993: 81n23),thiswouldcon-
stituteproofpositive of the archaicnatureof the phenomenonand its
independence fromtheClassical Latinaccent.
4.4.6. Anotherapparentcounterexample of thecauus type(4.4.2.) is
the verb lauere (lauï, lautus) 'bathe,wash (trans.)',wherethe a-vo-
calism in an initialsyllableagain appearsproblematic, at firstglance,

78Somewhat Driessen,forthei-stem,althoughhe allows forthepossibility


differently
of a source formlike *douH-ó-.For the morphologicalpattern-o- ►-et-,cf. the type
àpyóç 'shining' - ►àpyr|T-/àp78T- 'id.', etc. (Schindler 1976: 351); withö-graderoot,
e.p. Kofinoc'vouth' - ►kodotîtec'voune warriors'.
79Or "anakinetic",in
Tremblay's analysis(1996: 98f.).
8UFor the
antiquityof such denominativeformationsin *-ãjo/e-,cf. Ven. donasto
'gave'.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
s Law' inLatinandItalic
On 'Thurneysen-Havet' 239

no matterwhatkindof o-gradepresent{*lóuh3-o/e-, etc.) one wishesto


reconstruct.81 Beside thematiclauere, however,is the a-verb lauãre
(lauï, lautus),at firstwiththesame meaning,buteventuallyspecialized
withintransitive sense.This is also normallytakento derivefroman o-
grade formation, e.g. as denominativeto an abstractnoun *louh3-éh2
(thus LIV), or via an originaliterative*louh3-éio/e- (thus Rix 1999:
519).82 Those who prefer to take thematic lauere from an old e-grade
(thus both LIV and Rix) make the relativelystraightforward claim that
thevocalismof lauere is simplyreshapedafterthatof lauãre; and the
sameapproachis availableiflauere continuesan old ö-grade.
This gambitmaybe at leastsomewhatproblematic, in viewof thefact
thatlauere itselfhas impeccablyarchaiccredentials[Pit.,Cato, Enn.,
Naev.+]. Anotherexplanationcould begin withthe strikingprofusion
(and persistence)of x-luerecompoundswith preverb,which are far
morecommonthanlauere itself:thuse.g. abluere [Pacuv.+], alluere
[Lucr.+], dïluere [Pit., Cato+], ëluere [Pit., Cato+], prõluere [Plt.+];
attestedlater,also colluere[Ovid+], subluere[Caes.+], cf. MuerePrise.
Gramm.1.130.13 as the source of illuuiës [Ter.+]. At the same time,
with one trivialexception,83 lauãre makes no compounds.Thus an
alternativeto interpreting the vocalism of lauere as remodeledafter
lauãre would be to considerit the regularThurneysen-Havet' s Law
resultof originalö-gradeformswithpreverb,i.e. lauere <- ^x-laue-,
withfl-vocalism regularlyfrom*x-loue-(< ^x-louh^o/e-).Note thatin
this case, the initialaccent would reflectnot the pre-Latinperiodof
initialstress,but rathera survivalin Proto-Italieof the usual pattern
seen in Vedic Sanskrit(and normallyreconstructed at least fora late
of
stage Proto-Indo-European) in which verbs in main clauses show
unaccentedroot but accentedpreverb.An additionalsource of a-vo-
calism mightalso have been theverbaladjective,if thiswas originally
properto theiterative thatmayunderlielauãre - i.e. lautus< Haueto-<
(Transponat)*louh3-etó-(with *-eto-restoredfor *-oto-),cf. cautus
(3.2.).
4.4.7. A finalexampleof "unshiftedou" (cf. 2.2., 2.3.) has already
been mentioned(1.3.), namelythe modal formsof the shape /douï-/
meaning'give' (and, vis-à-visPIE *deh3-'id.', withunexpectedroot-
finalglide) seen in Fai. douiad, U. -tuvitu/ '-douitu,-tuvies,and (per-

81See thedetaileddiscussions
bySchrijver(1991:396f.)andMeiser(1998:85).
82 ver(1991:397),viastativeHaua-e-,basedon *laua-inlauere.
DifferentlySchrij
83Leumann (1977: 135),whootherwise takeslauerefrom an o-grade*louere,consi-
in Plautus
dersêluõ to derivefrom*êlauõ(to lauãre),givenëlauïlëlauisse(intrans.)
(4x);butthisis essentially
a kindofnonce-formation,
see Sommer1914:562.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
240 BrentVine

haps) OLat. duim,duat. The sourceof theseformsconstitutes a long-


standingproblem,84 thedetailsof whichare notessentialhere.Indeed,
theproblemof whethertheseformsbelongwitha root*deuH-'bestow'
(cf. IEW 218, 2. deu- '(religiös)verehren etc.') or (as in LIV) witha u-
extendedversionof *deh3-'give' (cf. on dautia,4.4.5.) can also be left
aside; the latterassumptionis adopted at the outset for expository
convenience(thoughcf.n. 85 below).
Lat. duim,if takenat face value, can reflecta zero-graderootaorist
optativestem *duhrihr (thus LIV), and in thatcase would pose no
problemforThurneysen-Havet's Law. It could well, however,recover
an Italic ograde stem *douï-(e.g. via a *-duul-generalizedfrompre-
fixedforms;see Leumannfordetails),and would thus,togetherwith
theFaliscan and Umbrianforms,need to be accountedfor.According
to LIV, thesereflecta full-grade ¿o/e-present "Neubildung"based on
<
therootaoriststem *dou- *déh3u-. so, If the pre-Italic*dóu-io/e-in
question(with/ouAdiphthong followedby consonant)would nothave
met the structural descriptionof Thurneysen-Havet's Law; and later
formswith/ouV/(^dóu-iio/e-,*dóu-iiã-,etc.) would no doubt have
retainedthe rootaccentintothe Italic period,hence remainingunsus-
ceptibleto Thurneysen-Havet's Law.85Accordingto a different account
(Jasanoff1991: 108nl5),86the Italic modal formspointingto *doui-
continuetheinheritedrootaoristoptativeof unenlarged*deh3-,which
in Jasanoffs reconstruction showedfullgradeof therootin 1 pl. and 2
pl. (thus *déh3-ihi-me,*déh3-ihrtë). For theresulting*dó-ì-mos,*dó-ì-
te (at firstwithhiatusbetweenrootand suffix),Jasanoffsuggeststhat
contraction would have been inhibitedby themorpheme boundary,and
thata hiatus-breaking w-glide would have been inserted betweenthe
rootand the mood sign (*dó(u)ì- > *dóui-).In thisevent,the attested
forms,withoutapplicationof Thurneysen-Havet's Law, could again
point to a in
retention Italicof root-accented *dóuì-,rendering suchma-
terialjust as unproblematic for the presenttheoryas in the LIV or
Driessenversionsof itsmorphological background.87

84 See Leumann 1977: 528, Meiser 1986: 186ff.,LIV s.v. *deh3u-,Driessen 2003:
355nl2. Here also Sicel dohit(Prosdocimi1998: 344)?
85 to a root *deuH- (as
is an analysisvia e-grade*déuH-j[o/e-
Equally unproblematic
favoredby Driessen,loc. cit.),sinceThurneysen-Havet's Law predates*eu > Ital. *ou.
80In effectan
elegantversionor a solutionproposedby Szemerenyi;see Leumanns
discussion(n. 84), withreferences.
87 Yet anotherrecentaccount that
operateswithplain *deh3-is thatof Rasmussen
(1999), who proposesa seriesof analogies (partlybased on fu-) thatwould again pose
no problemforThurneysen-Havet's Law.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' inLatinandItalic
On 'Thurneysen-Havet's 241

5. Lat. fauïlla: a new Thurneysen-Havet'sLaw etymology


5.1. One indexof the value of a new theoryis its abilityto generate
solutionsto hitherto recalcitrant
problems.Lat. fauïlla 'ashes, embers'
[Ter.+] is an extremecase, posingsevereetymologicaldifficulties. As
shownby Schrijver(1991: 442f.),thetraditional comparisonwith/ow^õ
'warm' cannot be accepted, whetherthe a-vocalism of fauïlla is
interpreted via Thurneysen-Havet's Law (whichshouldnotoperateon
the resultof *-ogyhV- > ^-ouV-) or in termsof a zero grade with"re-
duced vowel" (i.e. a *dheg"h-, impossibleto defend).As forthe suffix
formation: thelengthof the-I- is guaranteedbothby inscriptional spel-
ling with"I longa" (CIL V 3143) and by Romanceforms(see Meyer-
Liibke 1935: 280). Schrijversuggeststhatthispointsto a source like
*-i-Hn-leh2 (perhapsbased on an /-stem),notingfutherthat"[i]t is in
fact verydifficultto reconstruct an acceptable conglomerateof PIE.
suffixesthat would have yielded Lat. -ilia". He concludes that the
originof the formis unclear,and thatthe wordmaybe an "unknown
borrowing".
5.2. Thereis, however,an attractive rootetymology thatseems never
to have been considered,namelycomparisonwith the root *dheuh2-
(LIV1) or *dhueh2- (LIV2) 'Rauch machen'88 of Lat. suffire'fumigate',
fümus'smoke',fuligo 'soot', to cite onlyLatin material.Accordingto
this suggestion,fauïlla would constitutean isolated Thurneysen-Ha-
vet's Law form,withLat. prevocalicfauV-derivingfrom*dhouh2-V-.
5.3. Note further thatan originalroot shape *dheuh2- forthe verbal
root (as in LIV1) is not required.Like the primaryverbformations,89
most of the old-lookingnominal derivativesare zero grades, e.g.
*dhuh2-mó- (Lat. fümus, Ved. dhümá-, 9duóç, Oes dymü, etc.),
*dhuh2-li- (Lat. fuligo,Skt. dhüli-/dhüli- 'dust,soil, powder',Li. dülis
'smokescreen'[used in bee-keeping]),and otherisolatedand possibly
archaicitems(e.g. Oír. dé 'smoke',90Hitt.tuhhuwãi-ltuhhuwi-91 'smoke,
vapor', Toch. A twe/Btweye'dust'92).Thus, even foran originalroot
*dhueh2- (as in LIV2), one mightwell expectto find"wrong"fullgrades
- which in fact are attested,as in CS *dujp/duti'blow', Go. dauns
'sense of smell;fragrance'.Thus a basis *dhouh2- forLat.fauïlla raises
no seriousdifficulties in termsof rootstructure.

88Or(Watkins 2000: 19)'"toriseina cloud',as dust,vapor,orsmoke".


89Gk.3úco'sacrifice',
Lat.suffire,
etc.(see LIV).
90Watkins1966: 118n9,cf.more 1995:337 and(withotherrefe-
recently Schrijver
as wellas alternative
rences, conceptions oftherootshape)Vijünas2006: 132.
VlForthe from*deuh2-, seerecently Zeilfelder
2000:503f.,2001:254.
etymology
92Ringel996:52.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
242 BrentVine

5.4. Given the profile of the above material,involvinga root


etymologyforan isolatedform,one can do littlemorethanspeculate
aboutthemorphological and semanticdetailsof theformation.It is pos-
sible,nevertheless,to provideplausiblederivationalscenarios,based on
theassumptionthatan earlyItalic *pou-V-(< PIE *dhouh2-V-) has led
to Lat. *fauV-by an applicationof Thurneysen-Havet's Law. One pos-
sible startingpoint (cf. Schrijver'sremarkson the suffixformation,
5.1.) would be a primaryaerostatic/-stem*dhóuh2-i-, witha meaning
like 'smoke' or 'soot' (forexample,substantivized froman /-abstract
'smokiness').93But the remainderof the suffixformation need not in-
volve a segmentation *-i-Hn-(Schrijver,5.1.), witha formant like the
Hoffmann-suffix. The paradigmof such an /-stemwould includeinstr.
sg. *dhóuh2-ih1 'with smoke/soot',which would yield an early Italic
*pouï. As Nussbaum has shown (1996, 1998), instrumental singular
formsof this sort are apt to serve as the basis for derived "de-
instrumental" adjectives,fittedout withone or anotherof theproduc-
tiveadjectiveformants, especially*-no-and *-to-.94The former, in PIE
terms,is typicallyoxytonein adjectivalfunction,in bothprimaryand
secondaryformations. If thisaccentualfeaturewas retainedintoProto-
Italic,the derivationalsequence would be *póui 'withsmoke/soot'- >
*pouï-no-(adj.) 'smoky,sooty' > *pauìnó- (by Thurneysen-Havet's
Law).95The age of the subsequent/-formation is impossibleto deter-
mine,so thatthenotationwithItalic *p-,as opposedto "pre-Latin"*/-,
becomes arbitrary at this point.All thatremains,in any case, is the
further derivationvia nominalizing*-lo- (or diminutive*-elo-), ap-

93If,as is now
generallyassumed,Hitt.tuhhima-means 'panting,gasping' or thelike
(and not,as previously,'cryof pain [esp. duringbirth]')and belongswiththisroot(thus
e.g. Schrijver 1991: 233), it may preservean indirecttrace of just such an /-stem
*dhóuh2-i-, if the -ima- formationinvolvesa remodelingof an original*tuhhi-y and if
the unlenited-hh- (for expected *tuhi-)could be accounted for. As Craig Melchert
pointsout to me, the -hh-could be ascribedto theinfluenceof the verbtuhhãi-'pant',
and a parallel for the derivation ^tuhf}i-- ►tufyfyima- mightbe available in Hitt.
ekunima-'cold', which could involve a secondary¿-stem*ekuni-(for the type,see
Melchert1999) beside the adjectiveekuna-,with *ekuni-remodeledto ekunima-so as
to fittheexpandingclass of -ima-nouns.Accordingto Oettinger(2001: 463), tuhfyima-
is simplyderivedfromthe verb tufyhãi-, but this is not likely to be correct:tuhhãi-
inflectsas an -aizzi-verb,nota ¿/-verb,and thushas no allomorphsm'tuhhi(-)/''tuhhiya-.
(I am extremely gratefulto Craig Melchertforgeneroushelp withtheHittitematerialin
thisnote,as well as on Hitt.tufyfyuwai-Aufyhuwi-(5.3.).)
y4
Applicationsof thistheoryappearin Vine 1999, Peters1999.
95A concomitant difficult)is thatthe original
necessaryassumption(not particularly
/-stem*póu-i-itself(possiblyin competitionwith*f)ümó-fromthebeginning,and now
divorced,by virtueof its aberrantvocalism,fromits derivative*pauïno-) would have
been lost.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
s Law' inLatinandItalic
On 'Thurneysen-Havet' 243

parentlyspecialized as a feminine(and thus possiblywithcollective


function);hence *pauìnó- - » pre-Lat. *fauïn-(e)la '(small?) smoky/
sootystuff= 'ashes,embers'>fauïlla. Notealso thatthecomplexity of
the suffixalformation itselfcalls fora relativelyinvolvedderivational
history- a requirement certainlyfulfilledby an analysislike thatjust
offered.
The same resultcould be achievedusingotherderivationalsources.
Thus, if one begins withan o-gradethematic* dhouh2-o-'smoking'
(whetherconceived as a tóuoç or xouóç form),a "de-genitival"ad-
*
jective dhouh2-iH-nó- 'smoking(adj.)' (of the typeequo- 'horse' - ►
equïnus 'of a horse', etc.; Nussbaum,opp. citt.) will again regularly
yield Italic ^pauinó-.If,moreover,the starting pointwas a touóç form
*dhouh2-ó-, thiswould have a
produced *pauo- (witha-vocalism)from
the verystartof the derivationalchain,comparableto theothertouóç
formssuggestedabove (as in cauus, rauus,and the derivationalback-
groundsof saucius and dautia).

6. Conclusions
The resultsof thisstudy- someof themnecessarilytentative - can be
summarized as follows:
(i) By virtueof its relativechronology(as analyzedby Schrijver)and
thefactthatit mayhave operatedin U. sauitu(1.2., Vine 2004: 622ff.)
and perhapsVen. ho.s.tihauo.s.(4.4.1.), Thurneysen-Havet' s Law is
among theearliestinnovations of Proto-Italie
(and the
perhaps earliest).
The traditional accountin termsof theClassical Latin accentis there-
foreuntenable.
(ii) The law was conditionedby the remnantsof the PIE mobile
accentsystemin Proto-Italie(3.1., 4.2., 4.2.1.); its Latintracesare thus
comparableto otherLatin phenomenasuspectedof being conditioned
by a mobile accent thatpredatedthe CommonItalic periodof initial
stress(4.2.2.).
(iii) Thurneysen-Havet' s Law convertedunaccentedloi (in termsof
the presumedProto-Italiemobileaccent system)to /a/,in the position
beforea heterosyllabic /u/(4.I.). The restrictionof theprocessto hete-
rosyllabicposition can be correlatedwith general facts about the
phoneticsof lip rounding(Kaun 2004) and about feature-spreading
withinand acrosssyllableboundary.(See further (v) and (viii).)
(iv) The rule operatedon two majorcategoriesof formsin whicha
pretonicProto-Italie*o precededa heterosyllabic labial glide:

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
244 Brent
Vine

(a) Italicdescendantsof PIE iterati ve-causatives: U. sauitu(1.2.), Lat.


cauëre (1.1., 3.2.), fauëre (3.3.), auëre (3.4.), (?) lauãre (4.4.6., Rix
1999: 519) (< *kou-éio/e- etc.);96
(b) Italic descendantsof PIE tojióç forms:cauus (4.4.2.), rauus
(4.4.3.), saucius (4.4.4.) (?) dautia (4.4.5., Driessen 2003: 354f.), (?)
laus (4.4.5.), (l)fauüla (5.4.) (< *kou-ó-etc.).
Otherpossible pretoniccases include *lou-ã - > lauãre (4.4.6., LIV;
butcf.(a) above) and Ven. ho.s.tihauo.s.(4.4.1.).
(v) The unaccentedloi was not restricted to (first)pretonicposition
(4.2.). Possible of
examples Thurneysen-Havet's Law operatingin other
unaccentedpositionsincludethefollowing:
(a) secondpretonic
• verbaladjectivesin *-etó-,e.g. *kou-etó- > cautus(3.2.), Hou-etó->
lautus(4.4.6.);
• (factitive)*kou-ãió/é- > cauãre (4.4.2.);
• (de-instrumental or de-genitivaladjective) *pou-ï-no-- » fauìlla
(5.4.) (or as in (iv.b) above);
• (diminutive)^oui-ló- > auilla (4.4.1.).
(b) thirdpretonic
• (compound)*oui-(polkó- > aubulcus(4.4.1.)
• (compound)^oui-pol-ó-or ^oui-pa-ló-- » õpiliõ (4.4.I.);
• (denominative)Hou-ed-ã-ió/é- > laudare (4.4.5.).
(c) firstposttonic
• (preverb+ verb) *x-lou-o/e - > lauere (4.4.6.).
(vi) Non-application of Thurneysen-Havet's Law is regularforProto-
Italic accented*ó (cf. (ii), (iii) above), as in *dw/-'sheep' (2.3.2., 3.6.;
PIE aerostaticnoun),formsof (non-Lat.)*bóu-'cow' (2.3.1., 3.7.; PIE
aerostaticrootnoun),tójioç formsor substantives derivedby accent-re-
traction{*kóu-ã in Port,cova, Span, cueva etc.,4.4.2.), modal *dóu-i-,
*dóu-iiã-(4.4.7., if ultimately builton thefull-gradeactiverootaorist
to *dehi-'give').
(vii) The potentiallyseriousLatin exceptionsto Thurneysen-Havet's
Law (as formulated here) can involveearlier^-grades,directlyor indi-
rectly(nouãcula, 2.2.1.; cloãca/cloãre,2.2.2.; mouëre,2.2.3.), or are
otherwisenotprobative('touere, 2.2.4.).
(viii) The restrictionof the rule to sequences /o$uV/,with hete-
rosyllabicloi and lui, contrastswith the PIE source formsof many
Thurneysen-Havet cases, in which loi and lui were tautosyllabicby
virtueof a root-finallaryngeal- thus PIE */kou$hiéie-/ vs. Pr.-It.

96Reconstructions
hereand below provideProto-Italieinputforms.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' inLatinandItalic
On 'Thurneysen-Havet's 245

*/ko$uéie-/, etc.; thephonologization


of Thurneysen-Havet's Law may
be relatedto thissystematicchangein syllablestructure (4.I.)-97
(ix) The traditionaletymologyof Lat.fauilla 'ashes, embers'(:foueõ
'warm') cannotbe correct;the word may insteadbe a Thurneysen-
Havetformbased on PIE *dheuh2- or *dhueh2- 'Rauch machen'{:fümus
'smoke' etc.) (5.).98

References

Allen,W. S. 1978:VoxLatina2.Cambridge.
Bammesberger, A. 1986:Lateinisch lücere.Gioita64, 101-2.
Bendz,G. 1943:Caeliana: Textkritische undsprachliche Studienzu CaeliusAurelia-
nus.Lund/Leipzig.
CEG: Chronique d'etimologie grecque,(see httpV/perso.club-internet.fr/flo.blanc/CEG/
index.html)
Chantraine,P. 1933:La formation desnomsen Qrecancien.Paris.
Collinge,N. 1985:TheLawsofIndo-European. Amsterdam/Philadelphia.
Driessen,C. 2003: *h2é-h2us-o-,theProto-Indo-European termfor'gold'.JIES31:3-4.
347-62.
Eichner,H. andH. C. Luschiitzky (eds.). 1999:Compositiones Indogermanicae inMe-
moriam Jochem Schindler.Prague.
Ernout,A. 1905/6: Le parlerde Prénested'aprèslesinscriptions. MSL 13,293-349.
. 1909.Les éléments dialectauxdu vocabulaire latin.Paris.
Ernout-Meillet:A. E. and A. M., Dictionnaire étymologique de la languelatine:
Histoiredesmots5. Paris,1985.
EWAia:M. Mayrhofer, Etymologisches Wörterbuch des Altindoarischen, ErsterTeil:
ÄltereSprache. Heidelberg,1986-1996.
Flemming, E. 2005:A phonetically-based modelofphonological vowelreduction. MS
[http://web.
mit.edu/~flemming/www/paper/vowelred.pdfl.
Gaide,F. 1988:Les substantifsmasculins latinsen ... (i)o, ... (i)onis.Louvain/Paris.
GarciaRamon,J.L. 2000: Indoiranische Würzelpräsentia unditerativer Verbalcharak-
ter,inB. Forssman andR. Plath(eds.),Indoarisch, IranischunddieIndogermanistik,
119-131.Wiesbaden.
. 2004: Homerisch oüpoi'Kielfurchen, Landgräben' (*uoru-ó-'derZiehende')und

97 Law,wherethelengthening ofPIE *o in opensyllablesis


CompareBrugmann's
blockedin laryngeal-final
roots:thelaryngeal servesto closethesyllablesin question,
andtheseeffects surfaceprominently in thesametwocategories as withThurneysen-
Havet'sLaw,i.e. iterati
ve-causativesandtóuoç/touóç forms (see in thesetermsMayr-
hofer1986:148;forthenominal forms, Hajnal1994).
98Earlierversionsofthispaperwerepresented ata UCLA Program inIndo-European
Studies"RoundTable"meeting (Los Angeles;February 2004),at theworkshop "Lan-
guageandDialectin ArchaicItaly"(University ofCalifornia,Berkeley; March2004),
andbefore theDepartment ofLinguisticsat TokyoUniversity (October2004).I extend
mygratitude totheaudiencemembers atthoseoccasionsfortheirhelpful commentsand
discussion.
Warmthanks, areowedtoKazuhikoYoshida(KyotoUniversity),
especially,
whosegenerous andsupport
hospitality enabledmetowritethebulkofthispaperunder
idealconditionsduringmysabbaticaltimein Kyoto,and to Moss Pike(Program in
Indo-European UCLA),forhisindispensable
Studies, researchassistance.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
246 vine
Brent

épíxo 'ziehen', gr. óàkóç und î&kcû,Kétexpovund hom. KéXoai,in A. Hyllestedet al.
(eds.) 145-54.
Hackstein,O. 1993: Eine weiteregriechisch-tocharische Gleichung:Griechischnxrfem
undtocharischB pyãktsi.Gioita 70, 136-65.
. 2002: Uridg.*CH.CC > *C.CC. HS 115.1-22.
Hajnal, I. 1994: Das BrugmannscheGesetz in diachronerSicht und seine Gültigkeit
innerhalbderarischena-Stämme.HS 107, 194-221.
Hofmann-Szantyr 1965: J.B. H. and A. Z., LateinischeSyntaxundStilistik.Munich.
Horton-Smith, L. 1895: The establishment and extensionof thelaw of Thurneysenand
Havet.A/P 16. 444-67.
Hyllested,A. et al. (eds.). 2004: Per Aspera ad Asteriscos:Studia Indogermanicain
honoremJens Elmegârd Rasmussensexagenarii Idibus Martiis anno MMIV. Inns-
bruck.
IEW: J.Pokorny,Indogermanisches etymologisches Wörterbuch.Bern/Munich, 1959.
Jasanott,J.H. 1991: The ablautoí therootaonst optativein Proto-lndo-buropean. Mòò
52, 101-22.
. 2003: Hittiteand theIndo-EuropeanVerb.Oxford.
Juret, C. 1913: Dominanceet résistancedans la phonétiquelatine.Heidelberg.
Kaun, A. 2004: The typologyof roundingharmony,in B. Hayes et al. (eds.), Phoneti-
cally Based Phonology,87-116. Cambridge.
Kim,R. 2000: Reexaminingthehistoryof TocharianB 'ewe'. TIES 9.37-43.
Kimball,S. 1987: *H3 in Anatolian,in G. Cardonaand N. H. Zide (eds.), Festschriftfor
HenryHoeniQswaldon theOccasion ofhis Seventieth Birthday,185-192.Tübingen.
Klein,J. 2003: reviewof B. Forssmanand R. Plath(eds.), Indoarisch,Iranischunddie
Indogermanistik. Kratylos48.44-52.
Kortlandt, F. 1980: H2o and oH2. LP 23, 127-8.
Krasukhin,K. 2004: Archaicfeaturesoí Indo-buropeanword-tormation: ine ureek and
Old Indie typetóuoç - touóç in a PIE perspective,in J.Clackson and B. Olsen (eds.),
Indo-EuropeanWordFormation,119-138. Copenhagen.
Kretschmer, P. 1892: Indogermanische accent-und lautstudien.KZ 31.325-472.
Kühner-Stegmann 1966: R. K. and C. S., AusführlicheGrammatikder lateinischen
Sprache4,II: Satzlehre.Hannover.
Kümmel,M. J.2000: Das PerfektimIndoiranischen.Wiesbaden.
Leumann,M. 1977: LateinischeLaut- undFormenlehre.Munich.
Lindner,T. 2002: LateinischeKomposita:Morphologische,historischeund lexikalische
Studien.Innsbruck.
LIV: H. Rix (ed.), Lexikonder indogermanischen Verben.Wiesbaden,2001 (2nd ed.;
1sted. 1998).
Livingston,I. 1997: A LinguisticCommentaryon Livius Andronicus.Cornell Univ.
Ph.D. dissertation.
Lubotsky,A. 1990: La loi de Brugmannet *H3e-,in J. Kellens (ed.), La reconstruction
des laryngales,129-36. Paris.
Maniet,A. 1975: La phonétiquehistoriquedu latin dans le cadre des langues indo-
européennes5.Paris.
Mayrhofer, M. 1986: IndogermanischeGrammatik, 1/2:Lautlehre(SegmentalePhono-
logie des Indogermanischen). Heidelberg.
Meier-Brügger, M. 2003: Indo-EuropeanLinguistics.Berlin/NewYork.
Meiser,G. 1986: Lautgeschichteder umbrischen Sprache. Innsbruck.
. 1998: HistorischeLaut- undFormenlehreder lateinischenSprache.Darmstadt.
. 2003: VeniVidi Vici: Die Vorgeschichtedes lateinischenPerfektsystems.Munich.
Melchert,H. C. 1994. AnatolianHistoricalPhonology.Amsterdam/ Atlanta.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 247

. 1997: Denominativeverbs in Anatolian,in D. Disterheftet al. (eds.), Studies in


Honor of Jaan Puhvel (Part 1: AncientLanguages and Philology),131-8. Washing-
ton,D.C.
. 1999: Two problemsof Anatoliannominalderivation,in Eichnerand Luschiitsky
(eds.) 1999:365-75.
Meyer-Lübke, W. 1935: Romanischesetymologisches Wörterbuch. Heidelberg.
Nussbaum,A. 1996: Latinacëtum,acütus,aurïtus,auïtus: fourof a kind?Paperpresen-
tedat the 15thEast Coast Indo-EuropeanConference(Yale Univ.).
. 1998: More on "decasuative"nominalstemsin IE. Paper presentedat the 17thEast
Coast Indo-EuropeanConference(Univ. of NorthCarolina,Chapel Hill).
. 1999: *Jocidus: an account of the Latin adjectives in -idus, in Eichner and
Luschiitzky(eds.) 1999:377-419.
. 2003: A benign interpretation. Paper presentedat the 22nd East Coast Indo-
EuropeanConference(HarvardUniv.).
. 2004: Cool *-ëd-'The Latinfrigedoand GreekàÀ,yr|Ôû)v, tt|K8Ô(ûv,and pïyeôavoç
types.Paper presentedat the 23rd East Coast Indo-EuropeanConference(Virginia
PolytechnicInstitute).
Oettinger,N. 2001: Hethitisch-ima-oder: Wie ein Suffixaffektiv werdenkann,in G.
Wilhelm(ed.), Aktendes IV. Internationalen KongressesfürHethitologie,Würzburg,
4.-8. Oktober1999, 456-77. Wiesbaden(StBoT 45).
OLD: OxfordLatinDictionary,ed. P. Glare.Oxford,1968.
Olsen, B. A. 1999: The Noun in Biblical Armenian:Originand Word- Formation.Ber-
lin.
Panagl, O. 1999: Ein bukolischesProblem,in Eichner and Luschiitzky(eds.) 1999:
437-45.
Perrot,J. 1961: Les dérivésen -menet -mentum. Paris.
Peters,M. 1980: Untersuchungen zur Vertretung der indogermanischen Laryngaleim
Griechischen.Vienna.
. 1999: Gall(o-lat.) marcosior,in P. Anreiterand E. Jerem(eds.), Studia Celtica et
Indogermanica: Festschriftfür Wolf gang Meid zum 70. Geburtstag, 305-14.
Budapest.
Pinault,G.-J.1993: Grandeuret excès: Avatarsdu morphème"àya-" dans le lexique et
le discours.Revuede Philologie65, 195-218.
Prosdocimi,A. L. 1978: II venetico,in A. Prosdocimi(ed.), Lingue e dialettidell'Italia
antica (Popoli e Civiltàdell'Italia Antica6), 256-380. Rome.
. 1998: Appuntisul verbolatino(e) italico.Vili. Siculo: La 3a personasingolaredel
preterito (e l'italicitàdel Siculo), in L. Agostinianiet al. (eds.), do-ra-qepe-re: Studi
in memoriadi Adriana QuattordioMoreschini,333-46. Pisa/Roma.
Rasmussen,J.E.: reviewof Schnjver1991. Acta Linguistica26, 175-205.
. 1999: Miscellaneous problemsin Indo-Europeanlanguages VII: 50. Old Latin
duat,duit,in SelectedPapers on Indo-EuropeanLinguistics,2.649-50. Copenhagen.
Rau, J. 1998: PIE *uóidu-/*uéidu- and itsderivatives.Die Sprache 40:2.133-60.
Ribezzo, F. 1931: Due filoni di lingua mediterraneanella toponomasticaitaliana (I.
tirreno-mediterr. tauro 'monte'; IL tirreno-mediterr. nepo 'corso d'acqua'). RIGI 15,
151-61.
Rieken,E. 1999: Untersuchungen zur nominalenStammbildung des Hethitischen.
Wies-
baden.
Ringe, D. 1996: On the Chronologyof Sound Changes in Tocharian, vol. 1 (From
Proto-Indo-European to Proto-Tocharian). New Haven,CT.
Rix, H. 1992: HistorischeGrammatik des Griechischen2. Darmstadt.
. 1996: reviewof Schriiver1991. KratylosAX,153-63.
. 1999: Schwach charakterisierte lateinischePräsensstämmezu Set-Wurzeln mit
VollstufeI, in Eichnerand Luschützky(eds.) 1999:515-35.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
248 Brent
vine

Saussure, F. 1879: Mémoire sur le systèmeprimitifdes voyelles dans les langues


indo-européennes. Leipzig.
in A. MorpurgoDavies and W. Meid
Schindler,J. 1976: On the Greek type Í7t7i£i)ç;,
(eds.), Studies in Greek, Italie, and Indo-European LinguisticsOfferedto L. R.
Palmer,349-32. Innsbruck.
Schrijver,P. 1991: The Reflexesof the Proto-Indo-EuropeanLaryngeals in Latin.
Amsterdam/ Atlanta.
. 1995: Studiesin BritischCelticHistoricalPhonology.Amsterdam/ Atlanta.
Schuhmann,R. 2003: Zur deminutiven Funktiondes *-/o-Suffixesin Substantiva,in E.
Tichy et al. (eds.), IndogermanischesNomen: Derivation, Flexion und Ablaut,
219-230. Bremen.
Sihler,A. L. 1995: New ComparativeGrammarof Greekand Latin. New York/Oxford.
Sommer,F. 1914: Handbuchder lateinischenLaut- undFormenlehre.Heidelberg.
Sommer- Pfister1977: F. S. and R. P., Handbuch der lateinischenLaut- und Formen-
lehre,I: Einleitungund Lautlehre.Heidelberg.
Stevens,K. N. 1998: AcousticPhonetics.Cambridge,MA.
Stuart-Smith, J.2004: Phoneticsand Philology:Sound Change in Italic. Oxford.
Sturtevant, E. H. 1940: The PronunciationofGreekand Latin . Philadelphia.
Szemerényi,O. 195 1: Ein lateinischer Lautwandelõw > ãwi. KZ 70.5 1-76.
. 1989: An den Quellendes lateinischenWortschatzes. Innsbruck.
Thurneysen, R. 1946: A Grammarof Old Irish.Dublin.
TLL: ThesauruslinguaeLatinae.
Tremblay,X. 1996: Un nouveau typeapophoniquedes nomsathematiquessurhxauxde
l'indo-européen.BSL 91.97-145.
. 2003: La déclinaisondes nomsde parentéindo-européensen -ter-.Innsbruck.
Untermann, J.2000: Wörterbuch des Oskisch-Umbrischen. Heidelberg.
. 2002: Zu den verbadicendiim Oskisch-Umbrischen, in L. Sawicki and D. Shalev
(eds.), Donum Grammaticum:Studies in Latin and Celtic Linguisticsin Honour of
Hannah Rosen,377-84. Leuven/Paris/Sterling, VA.
Vijünas,A. 2006: TheIndo-EuropeanPrimaryt-stems.UCLA Ph.D. dissertation.
Vine, B. 1982: Indo-EuropeanVerbalFormationsin *-d-.HarvardUniv. Ph.D. disser-
tation
. 1993: StudiesinArchaicLatinInscriptions.Innsbruck.
in V. Ivanov and B. Vine (eds.), UCLA Indo-European
. 1999: Latin -inãreZ-inãri,
Studies,Vol. 1, 71-84. Los Angeles.
. 2004: New thoughtson an old curse (Tab. Ig. VIb 60/VIIa 49), in Hyllestedet al.
(eds.) 2004:615-626.
von Planta,R. 1892: Grammatik der oskisch-umbrischenDialekte,I. Straßburg.
Wächter,R. 2001: Non-AtticGreekVase Inscriptions.Oxford.
Wackernagel-Debrunner: J. W. and A. D., AltindischeGrammatik, II.2: Die Nominal-
suffixe.Göttingen,1954.
Waide-Hofmann:A. W. and J. B. H., Lateinischesetymologisches Wörterbuch. Heidel-
berg,1938-1954.
Warmington, E. H. 1940: Remainsof Old Latin IV: Archaic Inscriptions.Cambridge,
MA/London.
Watkins,C. 1966: The Indo-Europeanword for 'day' in Celtic and related topics.
Trivium1, 102-20.
. 1969 IndogermanischeGrammatikIII/l: Geschichteder indogermanischen Ver-
balflexion.Heidelberg.
. 1995: How to Kill a Dragon: Aspects of Indo-European Poetics. New York/
Oxford.
. 2000: TheAmericanHeritageDictionaryofIndo-EuropeanRoots . Boston.
Wells,J.C. 1982: AccentsofEnglish,Vol. 2 (The BritishIsles). Cambridge.

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions
Law' in Latinand Italic
On 'Thurneysen-Havet's 249

Yoshida, K. 2004: Anotheraerostaticnoun in Indo-European,in Studies in Anatolian


and Indo-EuropeanHistoricalLinguistics,194-9. Kyoto.
Zeilfelder,S. 2000: Heth. tuhhuessarund der Thymian,in M. Ofìtschand C. Zinko
in Graz. 497-508. Graz.
ieds.ì. 125 JahreIndogermanistik
. 2001: Archaismusund Ausgliederung:Studien zur sprachlichenStellung des
Hethitischen. Heidelberg.

Department of Classics Brent Vine


Program in IE Studies,UCLA
289B DoddHall
405 N. HilgardAve.,P.O. Box 951417
Los Angeles,CA 90095-1417
USA
e-mail:vine@humnet.ucla.edu

This content downloaded from 185.44.78.129 on Tue, 17 Jun 2014 14:46:09 PM


All use subject to JSTOR Terms and Conditions

You might also like