You are on page 1of 56

Celtic and Non-Celtic Divinities from Ancient

Hispania: Power, Daylight, Fertility, Water


Spirits and What They Can Tell Us about
Indo-European Morphology
Blanca María Prósper
Universidad de Salamanca
indoling@usal.es

This paper examines several hitherto neglected divinities


attested in Latin votive inscriptions, mostly found in central
Spain: Suttunius, Velonsa, Nevara, Duillae, Airo, but also in the
western regions: Ulisus Tereun(icus), Lumbae, Poltucea, Uladus
and Nituceus. An etymological note on Gaulish Sucellus is
included. These divine names are interesting from the point
of view of etymology but also pose some questions about
Indo-European and Celtic verb and noun inflection.

1. Introduction1
This study is devoted to a number of dedications to
different divinities found in central and northern Hispania
which have been paid little or no attention from the point
of view of etymology. I aim to provide a detailed linguistic
analysis that may contribute both to the understanding of
their possible function, which of course largely escapes us,
and to their dialectal ascription, and seek to throw some

1
This work has been financed by the Spanish Government (Project
MINECO FFI2012-03657: La antroponimia indígena indoeuropea de
Hispania: Estudio comparativo). The notation used here for Celtiberian
follows Villar (1997), where the traditional distinction <ß> vs. <s> is
replaced by <s> vs. <z>. For epigraphic forms in the Latin alphabet small
capitals are used. The reader is referred to the editions.; otherwise, the
missing details are found in Hispania Epigraphica (online edition, HEp),
Epigraphik-Datenbank Clauss-Slaby (EDCS) or Epigraphische Datenbank
Heidelberg (HD). Celtiberian texts follow where possible Untermann
(1997, quoted with the reference letter K.). Abbreviations: PN =
Personal name. EN = Ethnic name. FN = Family name. This work is
indebted to the patience and knowledge of Julián Méndez Dosuna and
two anonymous reviewers.

Volume 43, Number 1 & 2, Spring/Summer 2015


2 Blanca María Prósper

light on phonetics, word formation and analogical


processes in Hispano-Celtic.

2. SVTTVNIO (Autrigones). Fertility and lost laryngeals


SVTTVNIO DEO SAC(rum) / L(ucius) AVFIDIVS
MASCVLINVS / SESQVIPLICARIVS / P(ropria) P(ecunia)
FAC(iendum) CVR(avit), Poza de la Sal, Burgos, reading by
Abásolo Álvarez and Albertos Firmat (1976).
This form may have a cognate in the female PN in the
dative SVTTVNI in Noricum, in which all the geminates are
correctly spelled: DRIPPIO / LOTTAE F(ilio) ET / SVTTVNI /
AGRIPPAE / F(iliae) VXORI / INGENV (u) S F(ilius) V (ivus)
2
F(ecit) (Sankt Martin, Carinthia, Austria).
The divine name SVTTVNIO may be traced back to a
derivative of Celtic *sûtu- ‘fruit, product’, continued by OIr.
suth ‘fruit, offspring’ and consequently may be roughly
translated as ‘fruitful’, or ‘(divinity of) offspring’. On the
morphological side, it goes back to an IE action noun
*suH-tu- of the root *seuH-, which should be reflected in
Irish as †súth with a long root vowel to judge from Skt. sútu-
‘pregnancy’ (since AV.).
But what about the string <TT >? In the field of
epigraphically transmited onomastics, occasional doubling
of consonants is often put down to emotional causes and
labelled as ‘expressive’, or even as ‘inorganic’, which
conveys a pregnant meaning, ranging from ‘non
etymological’ to ‘unaccountable’. For instance, potter’s
names show a rather suspicious amount of structures of the
type CVCCV-. To complicate matters, inscriptions bear
testimony to such phenomena as misplacement of the
geminate (by which the scribe duplicates the wrong
consonant), simplified transcription of names with more
than one geminate, or omission of the second consonant
due to lack of space. Unfortunately, we are often unable to
detect original clusters which have undergone assimilation
processes Regional differences of course play a role: When

2
The text in EDCS-14400164 is incomplete. According to HD-057310,
the nominative corresponding to SVTTVNI would be Suttu. I believe
Suttunis or even Suttunia is more likely, in the latter case with
abbreviation for SVTTVNI(ae); alternatively it might be an indigenous
dative of the kind of Bhlhsami.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 3

rendering a name coming from a foreign area, a scribe may


make use of a double consonant to reflect a segment that
was a tense or aspirated stop in the original system. As a
consequence, we cannot expect uniform rendition of a
name everywhere.
Finally, resyllabification of a consonant preceding
heterosyllabic -i-/-u-, a typical resort intended to optimize
the syllable structure (and eventually leading to glide loss)
may account for several obscure geminates. This is the case
of some nasal stem PNs consistently showing doubling.
They are arguably derivatives from -i- or -u- stems: SATTO
(Narbonensis, Lugdunensis) from *sátu-ú ‘having plenty’,
cf. Lat. satur; VATTO (Narbonensis, Germania, Britannia)
from *uátu-ú ‘having poetic inspiration’, cf. VATVRVS
(Narbonensis); BVTTO (Noricum, Pannonia) from *butu-ú
(IE *bhu(h2)-tu-?), cf. BVTVNAE (Alpes Cottiae); MATTO
(Narbonensis, Noricum, Rhaetia) from *matu-ú ‘good?’, cf.
MATVNA, DEO MATVNO (Brittania); CRETTO (Belgica) from
*kreitu-ú ‘having discernment?’; LOTTO (Noricum) from
*floutu-ú ‘having riches?’, cf. LOTVRVS (Noricum); PATTO
(Germania) from *kuátu-ú ‘having motion?’, cf. PATTVO,
PATTVA, PATTOSVS, PAT( T) VRVS (Germania, Pannonia),
beside OIr. cáith ‘chaff’, Lat. quatio ‘shake’; BLATTONI
(Rhaetia), from *blátu-ú ‘flowering’, cf. OIr. bláth, BLATVNA
(Pannonia); BANNO (Venetia et Histria, Alpes Maritimae)
from *banu-ú ‘pig’; OTTO (Pannonia, Noricum) from
*foutu-ú ‘having fear’, cf. OIr. úath, OTVNI (Noricum);
BITTO from *bitu-ú ‘full of life’, beside BITVNVS (Belgica,
Germania, Transpadana); SASSONIS (Belgica) from *seh2t-tu-
ú or *seh2t-t-ú 'having satisfaction', beside SASVRA (once in
Pannonia;3 this is the obvious origin of the middle
participle SASOVNA in Delamarre 2004, from *seh2tt-o-
mno-).4 The IE noun *gueh2-tu- 'traversing, passing'5 is
attested in the nasal stem BATTONIS (Venetia et Histria),
beside BATVN[-] (Dalmatia) and three instances of BATVNA
in the leaden tags of Siscia (Pannonia Superior, cf.

3
Edition by Demicheli et alii (2012: 189).
4
A secondary stem in OIr. sásaid 'satisfies'. Cf. sás 'implement'.
5
Cf. the later, euphemistic ICelt. meaning 'death' in OIr. bath, etc., from
*g u Ô 2-tó-.

Volume 43, Number 1 & 2, Spring/Summer 2015


4 Blanca María Prósper

Radman-Livaja 2010: 189).6 Finally, LVTTONIVS (Ger-


mania), LVTTACI (Narbonensis) can be related to OIr. lúth
‘motion, vigor’, from *luH-tu-, but LVTVMARVS shows no
gemination; LVTTONIVS may obey the resyllabification rule
and go back to *lutu-ú. Conversely, METTVRICVM (Madrid)
and METTVRONI (Toledo) are derivatives from a word
*mettu- containing an unexplained geminate and probably
meaning 'shame' (cf. W. meth 'failure', OIr. metta 'shy') and
continued in continental onomastics (METTVS, Belgica,
Britannia, gen. sing. METTAI, Venetia et Histria).
The preceding evolution of the nasal stems is
suggestive of Continental Celtic sharing a feature well
attested in Celtiberian: consonant clusters were hetero-
syllabic, so that the expected syllabic structure of a
sequence like *matu-ú was *mat.u-ú, yielding *mat.tu-ú by
resyllabification, and finally *mat.t-ú with glide absorption.
In thematic formations, a sequence -t.uo- usually became
-t.tuo-. This is reminiscent of Romance evolutions like aqua
> acqua, quattuor > quattor in It. quattro, Sp. cuatro), futuere >
*futtere (see Prósper 2014a, drawing upon Vennemann
1988).
But what about thematic forms in -tuno-? While
derivatives in -ro- from -(t)u-stems are usually very ancient
formations (see below 8. on POLTVRICIVS), more recent,
inner-Celtic derivatives like SVTTVNIO are usually derived by
means of a suffix -no-: cf. Hispano-Celtic VISCVNOS,
BETVNIA, MATVNA, CRASTVNONIS, GVSTVNA. We may
consequently gather both nasal stems in a single original
paradigm: For instance, the different forms of MATTO vs.
MATVNA may be explained by way of paradigm split from a
holokinetic *matu-ú, gen. sing. *matu-n-ós, where
resyllabification and loss of the approximant destroyed the
identity of the suffix. Thereupon, the nominative gave rise
to a more uniform inflection with lengthened grade of the
suffix which imitates the Latin cognomina of the type
Cató, Catónis (but is largely independent from them in
such areas as pre-Roman Celtiberia) and shows the Latinate

6
To complete the picture, some PNs in -<TAVVS> may be related to -u-
stems if they go back to *-óu; for instance, the EN Batavi would preserve a
hypostasized form *gueh 2-tóu-o-, referring to those taking part in the
action.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 5

form of the suffix in Latin epigraphy; and the genitive


gave rise to a derived thematic inflection, e.g., Gaulish
*matunos, *matuní. Predictably, evidence for a sequence -
<TTVN >- is sparse, and mostly limited to PNs whose base
contains an expressive geminate, like CATTVNVS, ATTVNVS
and TATTVNVS, possibly derivatives of *katto-, *atta-,
*tatta/o- that imitate the above scheme or contaminated
reflexes of *katu-no-, *(f)á-tu-no- (< *peh3-tu- ‘drinking’ in
Lat. pótus).
Whether the secondary distribution of these two new
stems can be termed dialectal or not deserves further
study; but a cursory glance at the extant evidence turns
the scales in favor of this hypothesis: Hispano-Celtic hardly
shows any remains of paradigms based on -CCuú, whereas -
Cuno- is the norm. A caveat immediately imposes itself:
Hispano-Celtic has spread the long grade of the suffix to
the rest of the inflection (e.g., Melmu, Melmunos), and
consequently some PNs can be suspected of going back to
an ancient derivative -Cún-o-; this is to say, some stems of
the form -Cú, -Cúnós became thematic prior to
accomodation of the nasal suffix to Latin phonetics,
retaining their -ú- even when attested in Latin epigraphy.
Still, hardly any of them has the structure -C.uún- 
-C.Cuún-o- > -CCún-o- predicted for -u-stems; consequently,
they are probably mostly based on thematic, not -u-stems
(and then the derivative chain would be -Co-  -Cú, -Cúnos
 -Cún-o-), or on the original genitive singular of
holokinetic nasal derivatives of -(t)u- stems conveying
possessive value, which was (R)-(t)u-n-ós according to the
above hypothesis. This is why SVTTVNIO merits a word.
In this work I shall contend there is a sizable number
of instances in which gemination may indirectly be
reflective of etymological differences. Specifically, the
indigenous equivalent of the sporadic phenomenon
known as ‘inverse Compensatory Lengthening’ (or ‘Littera
rule’ after its paramount Latin example) may have been at
play here. To begin with, the Hispano-Celtic divine name
SVTTVNIO may have a geminate because it actually reflects a
phenomenon whereby etymological -VC- alternates with or
is replaced by -V̆CC-. Other likely Hispanic instances of this
phenomenon (or phenomena) are the PN MEDVTTVS

Volume 43, Number 1 & 2, Spring/Summer 2015


6 Blanca María Prósper

(Dalmatia, a Bracaraugustanus) and the FN MEDVTTICORVM


(Soria), from an agent noun *med-(d)út-, or TRIDALLVS
(León, CIL II, 5715) which in my view goes back to a dvigú
compound *tri-dálo- ‘having three parts, threefold’.7 A
similar alternation is attested in other languages, e. g. Skt.
sthúla- > Páli thúla- / thûlla- ‘big’.

Does Inverse Compensatory Lengthening occur in Celtic?


Compensatory Lengthening (CL) has received
divergent explanations in recent years. Its most common
instance is CV̆C > CV, which takes place mostly in languages
already possessing a short vs. long vowel contrast. Within
the frame of moraic phonology, Hayes’ classical account
(1989) neatly explains the change as a means of
conservation of the mora count. 8 By contrast, Kavitskaya
(2002) argues for listener-oriented change: compensatory
lengthening occurs if the consonant which is not readily
perceived (and therefore lost) has relatively long vocalic
transitions (such as a glide); the perception of the vowel as
long arises from the reinterpretation of the vocalic
transitions as vocalic length. This length is then
phonologised when the consonant is no longer perceived,
when the conditions for the cause of the length are not
perceptually recoverable. Most recently, Sen (2012) has
come up with an original view of compensatory
lengthening in Latin, based on the hypothesis that in this
language, counter to nearly universal expectations, short
vowels were phonetically longer in heavy than in light

7
Rather than to a compound allegedly meaning something like ‘very
blind’ containing an inexistent prefix *t®- (as posited by Delamarre 2003:
135). In this regard, we should bear in mind that OIr. dál ‘part, share’ also
means ‘division, tribe’, and consequently the compound conveys a
military or ethnic value like ‘having three tribes’. It may thus be
associated with VACORIA (León) ‘having two armies’ (< *duo-korio--) and
TRIDONIECV (Sasamón, Burgos) ‘having three persons’ (< *tri-(g)donio-).
Cf. Prósper (2013a: 191).
8
It does not explain, however, cases in which the mora count does not
need CL in order to be preserved, as in -VnCfric.C- > -VCfric.C-. His proposal
of extra-long, trimoraic syllables is somewhat circular (in so far as it
remains unsupported by evidence coming from metrics or accent) and
does not suffice to explain this exception, which is an areal feature
covering all the western groups of IE (Celtic, Germanic, Italic and such
minor languages as Lusitanian).

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 7

syllables, which explains why they were more resistant to


abbreviation at the time of prehistoric initial accent. In
languages where closed syllable vowels are phonetically
longer than those in open syllables, compensatory
lengthening would be supported further by the
phonologization of the unexpected length after coda loss.
As regards ‘inverse CL’ in Latin, it is typically attested
in clusters of high vowels + voiceless stops and open vowels
before sonorants. Sen observes that high vowels are
phonetically the shortest in duration, and voiceless stops
are aptest to gemination, given the absence of any
aerodynamic difficulty. If vowels in closed internal syllables
were phonetically longer, when the phonetically shortest
long vowels (high vowels) were realised in the phonetically
shortest environment (before voiceless obstruents), they
were most susceptible to being reanalysed as short vowels
in closed syllables, since such a short vowel would be
roughly equal in length to the phonologically long open
syllable vowel. Consequently, it was not only the desire to
preserve syllable weight that led to the gemination of the
consonant, but also the perception of the vowel as a short
one in a closed syllable. The segment causing the closure
would be the following consonant, which was therefore
realised as a geminate (coda + onset). (Note that under
these premises inverse CL would hardly be expected to
take place in any language where closed syllable vowels
have the expected length.) The sequence ‘/a/ + sonorant’
leads to synchronic variation between -VC- and -V̆CC-, and
the evidence from Latin bears this out, e.g. flamma vs.
flama ‘flame’. Since low vowels have the longest intrinsic
duration and vowels are phonetically longer before
sonorants than before voiceless obstruents, /a/ and /a:/
were perceptually confusable in this context. According to
Weiss (2010), high vowels before voiceless stops were
reanalyzed as diphthongs. Glides are usually more
peripheral in this context: Then, the hyper-peripheral
glide was assimilated to the following stop: IIT > IJT > ITT.
Voiceless geminates are more natural than voiced ones,
which explains the restriction of this change.
It is not my intention to unravel this complex
problem, but only to deal with it descriptively, considering
geminates when necessary as indicative of a preceding

Volume 43, Number 1 & 2, Spring/Summer 2015


8 Blanca María Prósper

etymological long vowel. Bear in mind we really do not


know the extent to which either scribes or bearers of the
attested names, or both, were bilingual, and consequently
whether this phenomenon can be labelled as: a) one of
interference with Latin, b) one of usage of fashionable
indigenous names by speakers of Latin with at best a
residual knowledge of Celtiberian, or c) the way actual
speakers of a Celtic language rendered sequences which ex
hypothesi would have undergone the same ‘inverse CL’
process as in Latin. 9 In any event, it is an irregular and
structure preserving change, like metathesis or
dissimilation, not a regular sound change.
Why -V̆CC- has become the norm and excluded the
alternative in some cases cannot be ascertained, and Latin
bears testimony of the possibility of alternative
pronunciations for some words, to the point of eventual
split and semantic or morphosyntactic specialization, as in
cuppa/cúpa (Sp. copa ‘glass’ vs. cuba ‘barrel’), mútus/muttíre.
This is the explanation given to some Latin participles
which no longer belong to a verbal paradigm, such as glittus
‘sticky’ and uitta ‘woollen band’ as opposed to the easily
segmentable solútus, nutrítus, where only recognizable
morphological boundaries prevent the change. Other
words simply fail to show any trace of it, like lítus ‘coast’ >
It. lido, nítor ‘to lean on’, ícó ‘to strike’. We should keep in
mind that the Latin input for ‘inverse CL’ is somewhat
larger than the Celtic one, since it includes original
diphtongs.
It would take us too far afield to investigate whether
‘inverse CL’ is shared by other Celtic languages, or even
can be traced back to Common Celtic, at least as a
tendency. This is far from being a regular change, and it
may have occurred first in slovenly or lower class speech,
and then become standard only in particular cases (as some
9
In any event, the geminated -tt- in SVTTVNIO militates against the
widespread, but unfounded idea that intervocalic voiceless consonants
became voiced in this area of Hispano-Celtic. I am presently preparing a
geographically finegrained work on indigenous anthroponymy in
Celtiberia proper, where I seek to show that there is not a single
example of intervocalic voicing and that, by contrast, voicing of stops in
the neighbourhood of a nasal segment is universal except in the
southwestern areas.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 9

epigraphic evidence with a geminate vs. the outcome in


the Romance languages would seem to suggest). The
considerable differences in the transmission of
Continental vs. Insular Celtic languages and their disparate
chronologies are consequently conducive to biased,
incomplete and, in sum, unreliable results. A last word must
be said about our inadequate knowledge of Continental
Celtic accent. The present hypothesis can be sustained for
some names at the cost of accepting that the input for the
inverse CL is provided by allophonic lengthening of
accented vowels in trochaic structures, so that CV CV >
CVCV > CV̆CCV, which might be held responsible for the
large number of Gaulish PNs of the form C(R)V̆CCV-. Since
all this ultimately hinges on statistical plausibility, in the
following lines I shall especially regard those names which
are not so widely attested as to be suspect of being just
Latin, and where evidence of alternation of a simple vs. a
double stop is not clear. Suffice it to say that several
examples speak in favor of this possibility:
CCelt. *sukko- ‘piglet’ can be reasonably explained as
the outcome of *súH-ko- (cf. Lat. succidia/súcidia ‘joint of
pork’)10 with the suffix -ko- which is widespread in animals’
names. In fact, this is exactly the form inherited by Vedic
súkará- = Lat. súcula (but epigraphic SVCCVLA once), Germ.
*súgó. 11
*bukko- ‘goat’ is not necessarily a loanword from
Germanic or a derivative *bug-ko-, since Avestic has búza-
with a long vowel (often taken to be the original form) as
opposed to ‘expressive’ Skt. bukka-, which points to a
primitive alternation.
*krittá ‘body, frame, shape’, in MIr. crett, creit, MW.
creth ‘nature, appearance, form’ is likely to be a noun of
the ‘uitá’ type, from *kriH-téh2 or *krHi-téh2 with laryngeal
metathesis, or the fossilized feminine of a past participle of
the same structure. A good candidate is the root *kreiH-
‘sich auszeichnen, vortrefflich sein’ (Rix et alii 2001: 337).
*krutto- ‘round object, womb’, in OIr. crot ‘harp, lute’,

10
Note a variant *su- does exist and probably originated in the gen. sing.
*suH-ós.
11
This form may have influenced the synonymous *mokko/u- (< *mokó-?)
and not the other way round as often assumed.

Volume 43, Number 1 & 2, Spring/Summer 2015


10 Blanca María Prósper

MW. crwth ‘round object, harp’, croth (fem.) ‘womb’ can be


explained as a past participle *kruH-tó- (or *krHu-tó- with
laryngeal metathesis), which Matasovic (2009: 228)
hesitatingly compares to Lith. krútìs ‘woman’s breast’. Cf.
CRVTTO (La Graufesenque), CRVTONIVS (Venetia et
Histria).12
*brikko- in OIr. brecc ‘speckled’, Gaul. BRICCO, etc. has
no convincing etymology thus far. Pace Matasovic (2009:
78) its initial labial cannot go back to IE /p/.
OIr. cittach, cettach ‘lefthanded’, MW. chwith is
reconstructed as *skh2i-tto- with ‘expressive gemination’
with some scepticism by Matasovic (2009: 343). Perhaps we
should depart from a double pronunciation *skíto-/*skitto-.
For many other words we find unequivocal evidence
of gemination only in Continental Celtic, and specifically
in Gaulish PNs (note that <O>, < E > occasionally occurs in
lieu of ‘correct’ <V >, <I>, as in the prefix su-, so-, etc.):
MIr. crích ‘boundary’ (< *krí-kuá) has a likely cognate
in the Celtic PN CRIPPO (< *krikukuú, Transpadana,
Germania, Venetia et Histria).
The exact equivalent of Lat. brútus (< *gu®h2u-tó-
‘heavy’ with allegedly Sabellic Anlaut) is unattested in
Insular Celtic, but two instances of BRVTTA in Noricum (in
inscriptions where all the names are otherwise Celtic), and
one BRVTTAE in Rhaetia (an indigenous woman who casts a
curse on four men bearing Latin names and addresses it to
Ogmios) as opposed to one BRVTVLLA (Pannonia Inferior)
are diagnostic of the same phenomenon (although once
again *brutuá, a derivative of *brutu- ‘fermentation’ cannot
be ruled out). 13
Several cases of TITTVS 14 (Britannia, Belgica) in Celtic
territories and onomastic contexts, including prefixed
AT( T)ITONIS, ATITTO (Noricum) and ARETITONI (Lusitania

12
Matasovic’s conclusion on these forms, which resorts to an unknown
non-IE source, can be dispensed with.
13
Note that the gentilic Bruttius owes its geminate to context-bound
resyllabification before yod. Bruttus is unattested in Latin: the only
testimony would be CIL II, 3048 L(ucius) BRVT(tus) and TI(berius)
BRVT(tus) (Complutum), where I read BRVT(tius).
14
And TIT( T)ANVS, the gentilic Titanius, TITTAVSVS, TITTONI, TITTONIAE, a
rare gentilic Titonius, TIT( T)VLA, TITTVLONI, TITVLO (Narbonensis,
Rhaetia, Tarraconensis, etc.).

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 11

Emeritensis) may go back to a participle *tih1-tó- ‘heated,


molten’.15 DEOTITANO (Rhaetia) comes from *deiuo-títo-
and must mean something like ‘ìnflamed by the divinities’.
TICCO, TICCONI (Venetia et Histria) and TICCONIAE
(Burgos) may be related.
MICCO, MECCO, etc. (Germania, Belgica) may come
from *mh1i-ko-, a synonym of Lat. mítis, OCS. milû, OIr. mín,
‘meek, soft’, showing a variety of suffixes.
LICCVS, LICCA, LICCO (Pannonia, Dacia, Moesia) may
go back to *h2leh1-ko- ‘miller’.
DICCONIS (Moesia Superior) may be related to MW. dic
‘wrath’, Russ. dikij ‘wild’ from *díko-.
An isolated MELVTTA (Pannonia) is an arguable
cognate of the Hispanic MEDVTTVS with the common
misperception of [l] for [d].
AVTOSCVTTVS (Noricum), SEMVSSCVTTVS (Pannonia),
SCOTTVS (Belgica) has a near cognate in á-prati-skuta-
‘undisturbed’ (ani†) and its second member goes back to
*skuh2-tó- ‘disturbed’ (Rix et alii 2001: 561). Scheungraber
(2011) translates AVTOSCVTTVS as a possessive *auito-skuto-
‘Wunschklinge habend’ (cf. OIr. scoth ‘point’), but a
synthetic compound meaning ‘spurred by desire’ looks
more convincing to me.
PITTA, PITTVS (Venetia et Histria, Rhaetia, Moesia
Superior) is identical to Lat. quietus, Av. siyáta, from *kuieh1-
tó-‘quiet’ or *kuih1-tó- (*kui-tó- ‘white’ cannot be excluded,
however).
A Latin appellative iutta ‘soup’ is held to go back to
Gaulish. The double consonant is indicative of a preceding
long vowel, and 'inverse CL' may be either Gaulish or
Latin. Accordingly, this form is the direct continuant of
*(H)iuH-tV-, and its cognates MW. iwt 'porridge', perhaps
Gaul. IVTV-, etc. have undergone shortening (see Zair
2012, 139 on this set of forms). The doubling of the
consonant in iutta, however, might be the product of
transmission to Latin if an adapted *iutu-á had undergone
resyllabification, by which *iut.tu-á, and eventual loss of

15
cf. ON. pídr ‘thawed’, Lat. títió ‘burning brand’, perhaps OIr. úathad ‘a
small number, few’, MW. odit ‘exceptional thing’, from *au-títo-, the
original past participle of *ti-n-h 1- in OIr. tinaid ‘melts away’ (see Zair
2012: 117).

Volume 43, Number 1 & 2, Spring/Summer 2015


12 Blanca María Prósper

the approximant.
Gaul. BVDDVTTON ‘kiss/penis?’ on the spindlewhorl of
Saint-Révérien may go back to a deinstrumental possessive
derivative of a noun in -tu- of the type acútus16 and
consequently reflect *bussú-to- from bussu- ‘lip’, meaning
that a kiss abides in the lips. Nevertheless, if the base
means ‘penis’, the derivation is irrational unless, as has
been suggested before (see Delamarre 2003, s.u.), this
word actually designates the spindlewhorl as ‘penis-shaped’.
A similar case would be VAMNITTVS (Baralle, Belgica), an
adjective of the type facétus built to the lost root-stem
*mén- (but cf. *m÷-dh h1-ónt- in the Gk. aorist part. may≈n),
and accordingly going back to PGaul. *uo-mníto-, (as if)
from *h1upo-mneh1-to- ‘having little wits’ corresponding to
the Balto-Slavic infinitive *minéti, in turn based on a
‘stative stem’ *m÷(n)è-, and entirely as predicted by
Jasanoff's model (2003, 154-157).17
Finally, Gaulish ‘inverse CL’ clarifies a number of
geminates <PP> which in principle must go back to
labiovelars: SVPPO, SVPPONIS (Germania Inferior), SVPPA
(Pannonia Superior), SVPPVTVS (Belgica) may go back to
*súkuo-, in turn from an archaic compound *h1su-h3kuo-
‘looking good’. TALVPPA (Germania, Aquitania, Belgica,
etc.) may go back to *talúkuo- from *tÒHu-h3kuo-. 18
Interestingly, this reveals a new, hitherto neglected
numeral-based onomastic series crucially containing the
usual compositional form of the cardinal: I believe VIPPONI,

16
As established by Nussbaum in an unpublished but influential study of
1996.
17
The suffix -itto- may consequently be the deinstrumental adjective in
-eh1-to- of the stative -eh 1- verbs, originally linked to them as in Balto-
Slavic. It was somewhere down the line reanalyzed as enlarging a
thematic form, which may have given rise to VEPITTA, GLARITTA, PVSSITTA,
ALVITTA, etc. The same probably applies to names suffixed by -< ICC >- and
not -<IC >-.
18
Perhaps a near synonym with Skt. pratíka- ‘face’ (‘looking towards the
front’ or an ¶nyeow type ‘having eyes in front’ rather than a prepositional
‘in front of the eyes’), and then perhaps replacing the inherited *proti-
h 3k uo-. Gk. prÒsvpon, Toch. A pratsak, B pratsáko may come from an early
refashioned *protio-h 3k u-o- instead of a variant *proti-h 3 óku-o- or *protih 3 k u-o-
obeying the Francis-Normier rule. Cf. Skt. madhúka- ‘a plant like honey’,
svitíc ‘looking white’ (< *kuiti-).

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 13

VIPPIVS (Liguria, Alpes, Narbonensis) go back to *duíkuo-


‘double looking, twofold’; TRIPPI, TRIPPONIS, TREPPONIS
(Transpadana, Pannonia) to *tríkuo- ‘threefold’,19 and
under this interpretation a PN TRVPPICVS attested only in
Venetia et Histria must be traced back to *kutrúkuo-
‘fourfold’, containing a very ancient zero grade variant of
the compositional form *kuetru- in Gaul. petru-, Lat. quadru-,
etc.20 The hápax ELOPPO (dat. sing., Belgica) accordingly
goes back to *pelu-h3kuo- ‘manyfold’.
This compositional scheme may underlie a couple of
indigenous Gaulish PNs which usually show a geminate
<PP>: TAPPO/TAPPVS (Aquitania, Transpadana) and
IAPPOS/IAPPVS (potter's name) can be unified under a
single structure under the assumption that they go back to
(derogatory?) *to-h3kuo- and *io-h3kuo-, as in OIr. cách
‘everyone’, OCS. kakû, takû from *kuo-h3kuo-, *to-h3 kuo-. The
stop would be doubled after a long vowel, though not the
expected one. Compounds in -apo- from *h3ókuo- show no
gemination (LICCAPVS, ACAPVS, SACRAPVS) except perhaps
VRAPPVS (Narbonensis) and VRAPPIAE (Transpadana). This a
hitherto overlooked but arguably ancient PN: if it goes
back to *h1ur(u)-h3 ókuo- ‘wide looking’, it is identical to Gk.
eÈrvpÒ (cf. also eÈrÊopa, an epithet of Zeus ‘far seeing’),
and was replaced in Skt. by urucak§a- with active sense ‘far
seeing’.
This poses a new problem, however. Laryngeals are
commonly held to be lost in an originally pretonic syllable
in Celtic obeying Dybo’s rule (see Zair 2012 with a critical
account of former ideas). Since the exact conditions for its
action are not known and it has many exceptions, I have
not taken it into account. It would be impossible to review
here the inherently contradictory evidence for and against
different formulations of this rule, let alone the partially

19
The family of OHG. zwíg ‘twig’ might come from *duík uó- with labial
dissimilation. Cf. Skt. dvíka- ‘twofold’, Skt. tríka- ‘triad, threefold’ (with a
short vowel, possibly by analogy with other dvi- compounds); for both
families a velar is usually reconstructed.
20
Cf. the Venetic divinity in the dat. sing. trumusijatei (Làgole di Calalzo,
Cadore). The dialectal ascription of both forms is consequently doubtful.
Its best parallel is Gk. tru-fãleia ‘helmet’. On these forms, see Prósper
(forthc.-b).

Volume 43, Number 1 & 2, Spring/Summer 2015


14 Blanca María Prósper

conflicting action of the ‘Wetter rule’, which posits


laryngeal loss or vowel shortening preceding two
consonants and is complementary to the former.21 This
alternative traces back OIr. ith ‘corn’ to an allomorph *fî-
tuV- vs. íth ‘fat’ from *fí-tu- (< *piH-, cf. Widmer 2004: 19),
and perhaps bith ‘life’ to *bî-tuV- (< *guih3-) as well as our
above example OIr. suth to *sû-tuV- vs. SVTTVNIO, SVTTVNI
from *sú-tu- (< *seuH-) and perhaps LVTV- beside OIr. lúth.
These allomorphs may originally belong to derived forms or
have been abstracted from the instrumental *-tu-eh1. But
the discrepancy could also be reflective of the older Ablaut,
which required elimination of the laryngeal in the full
grade forms *peiH-tu-, *gueih3-tu-, *seuH-tu-.
In sum, after Celtic loss of the laryngeals, some of the
propitious contexts for ‘inverse CL’, specifically those with
a complex onset Cri/u:C- and sCi/u:C-, show traces of a
pronunciation [iC:], [uC:], except under paradigmatic
pressure (for instance, when the form is synchronically a -
to- participle). The process may have been reenacted in
the separate languages, and Latin influence cannot be
ruled out for Continental Celtic.

3. VELONSAE (Autrigones). A strong-willed divinity?


VELONSAE / CALPVRNIA / PATERNA SE/VERI F(ilia)
V (otum) S(olvit) / L(ibens) M (erito), Ranera, Tobalina,
northeastern Burgos.22

21
If we limit ourselves to the sequence -i/uHCV-, where C is a stop, the
clearest example of pretonic shortening is IE *b huH-tó- > PCelt. *bú-tó- >
CCelt. *büto- > OIr. ro-both ‘was’. The stem variant *bü- may have been
generalized early from the prevocalic position, however: the matching
Lat. fütúrus can hardly be the product of laryngeal loss, since Umbrian has
fito (acc. sing. n.) from *b hú-tó-. MIr. sith- ‘long’ comes from *sHi-tV- with
no metathesis. Counterexamples, in turn, show no shortening or had an
accented root: ro-bíth ‘he was struck’ < *bító-, ro-críth ‘he was bought’ <
*k urító-; MIr. drúth (adj.) ‘wanton, unchaste’, Gaul. Drutos < *drúto-; MIr.
cích ‘breast’ < *kí-ká; MIr. fíthe ‘woven’ < *uí-t-ió-; OIr. túth ‘preservation’
comes from *tú-tV-; MW. rhidiaf ‘copulate’ < *rítV-; MW. gwit, guid
(masc.) ‘feast, banquet, liquid, honey’ < *uítV-; MIr. níth < *niH-tu- ‘fight’;
líth < *liH-tu- ‘feast’. The obvious exception is OIr. moth ‘penis’ (= Lat.
mútó/muttó from *müto-), arguably influenced by toth ‘female genitals’.
22
This inscription is now preserved in the Museum of Navarra, in
Pamplona. It was originally edited with an erroneous transcription of the
divine name by Elorza and Abásolo (1974); the correct text was finally

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 15

The structure of this hitherto overlooked divine name


is transparent: We are dealing with an active participle in
-nt- whose synchronic structure was *uelont-iá, in turn
probably from a primitive *uel(H)-ont-ih2 that should have
given a Celtic form *uelontí. The symbol <S > can only be
accounted for by accepting that affrication of a cluster -ti-
took place in all possible contexts in that particular region,
at least as long as the word can be considered Celtic.23
At first sight, one is tempted to relate this word to the
root *uel(H)- ‘to command, have power over’ (*uelH- in Rix
et alii 2001: 676-677; Pokorny 1959: 1111-12), attested in
Lat. ualeó ‘be strong’, with an active part. ualens. Velonsa
would be suitably described as a ruling, powerful goddess.
This faces us with some difficulties, however, if VELONSAE
is a Celtic divinity, as would be in principle expected in
that area: There is an increasing number of names based
on participles of this root attested all over Europe, and
they consistently show deponent inflection (like the
present formation preserved in OIr. follnathar): VALAVNA,
Lep. Ualaunal may be middle participles in -mno- or
thematized agent nouns based on the original genitive
*uÒH-mn-os with a secondary full grade of the root.24 British
and Gaulish *uellauno- in PNs and ENs like Cassivellaunus
may go back to a thematized, secondary agent noun based
on a present stem *ualnamú (cf. its derivative OIr.
follamnacht ‘government’) or to a middle participle

established by Castillo, Gómez-Pantoja and Mauleón (1981, nr. 28, 55).


23
See Prósper (2014a) for the Celtiberian alternation -<nti>-/-<NTI >- vs.
-<nz>-/-<NS>-; the latter exhibits a far more restricted geographical
distribution than when -t- is preceded by a vowel, showing that -i- was
prone to vocalize in the ‘Sievers’ context, checking the tendency of the
cluster to affrication; -C.ti- yields -C.tii- at least in central Celtiberia.
24
The athematic formation is preserved in my view in the PN (acc.)
VALAONEM, recorded as the chieftain of the Germanic Naristae (Suebi) in
the column of Marcus Aurelius and AE 1956, 124 (Numidia). Since the
original nom. was *ualamú and the acc. was *ualamon-am, this could mean
that the suffix had been refashioned in Gaulish after the oblique cases
like the genitive *ulaunós. Therefore, the nom. sing. that the Romans
actually heard, and adapted to their inflectional patterns as *ualaó,
*ualaónis, must have been *ualaú. Other Suebic chieftains, such as
Ariovistus and Ariogaesus equally bear Celtic names. The corresponding
feminine form is attested in Celtib. launi (< *uÒh 1-mn-ih 2, cf. Prósper
2014a).

Volume 43, Number 1 & 2, Spring/Summer 2015


16 Blanca María Prósper

*uelnamno- showing a secondary full grade of the root. 25


Velonsa would have to be a secondary aorist form, the best
parallels of which would be Toch. Bwalo, Awäl, gen. lánte
‘king’, from *uÒH-ónt-, *ulÓ-nt-. The original athematic
nom. *uelÓ-nt-s is attested in the first term of the
Tocharian compound Bylaiñäkte, Awlámñkät ‘divine king’
(cf. Lubotsky 1994). If they were directly represented in
Hispano-Celtic, they would have yielded respectively
*ualontiá and *uelantiá.
Other etymological possibilities are open: An IE root
*uelh1- meaning ‘to want, choose’ (cf. Rix et alii 2001: 677-
678, Pokorny 1959: 1137) forms the base of Lat. uoló,
which is an original athematic root present, which, if
unchanged, would end up in our case as *uelansá from a
refashioned nom. *uelh1-÷t-ieh2.26 Yet, Lat. uoluntás ‘will’
may be traced back to a thematic participle of this verb,
originally *uel(H)-ont-, and a divine name Herentas with a
similar meaning is found in Paelignian (a dialect of the
northern Oscan group) and traced back to *gh er-÷t-; this
goddess is held to be the equivalent of Venus on the
testimony of a Hesychian gloss reading ÑEri°nthw:
ÉAfrod¤thw §p≈numon. Nussbaum (1976) regards *uel(H)-ont-
á- as preferable to the traditional reconstruction *uel(H)-
ont-tát-. In fact, a gentilic VOLVNTILIVS is directly derived
from the thematic participle. In sum, both the
reconstructed Latin participle *uelont-, later refashioned as
an abstract, and the probably Hispano-Celtic divine name,
could be relics of an ancient, presumably common
language stage.
Finally, VELONSAE might go back to a root *uel- ‘to
see’ (Rix et alii 2001: 674, Pokorny 1959: 1136-1137),
which has the obvious advantage of having a simple,
thematic root present *uel-o/e- that is directly attested in
OIr. fil ‘there is’ but has seemingly been resuffixed in
Proto-Brittonic (see Schumacher 2004: 669-671); it is
semantically related to prophetic vision in the agent noun

25
The original form *ualnamno- is preserved in VALLAVNA, VALLAVNO
(Pannonia Superior), VALLAVNIVS (Britannia).
26
Joseph’s law does not operate in an original sequence -eC÷-. Gaul. VELOR
(Marcellus of Bordeaux) has been identified as a dep. 1st sing. present of
this root (Fleuriot 1974: 60).

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 17

OIr. fili, filed, Ogamic VELITAS, reflecting *uel-ít- ‘seer’


(ultimately *uel-ét-; see Delamarre 2003: 311 for
Continental Celtic cognates).

4. CALECAE NEVARAE (Pelendones)


A group of votive inscriptions from Villaverde del
Monte (central Burgos) reads:27

1) AIVS / CALECE / NEVARE / V(otum) L(ibens) [A(nimo)]


S(olvit)
2) CARECE N/EVARE · AV/A LECIRA · P/RO · S(alute) ·
V(otum) · S(olvit) · M(erito)
3) CALECAE · / NEVARE · / V(otum) L(ibens) S(olvit) /
LICINIVS · STRENVVS
4) CAL[---] / ----

The structure of these dedications consists, as usual in


western regions, of a divine name plus an epithet. But in
this particular case, Caleca is furnished with a velar suffix
and is likely to designate the protector of a village or
region. Since this group uniformly shows <E > in lieu of
<AE >, we may conclude that it renders a dative *kalaikái.
This name is followed by an epithet *neuarái,28 which looks
like a continuant of the IE adjective meaning ‘young, new’
in Gk. nearÒw, Arm. nor ‘new’29 and Lat. nouerca
‘stepmother’30 . Wodtko et aliae (2008: 525) reconstruct
*neuero- with a question mark. All things considered, the
attested forms point to *neuÔ2-ro-. The occurrence of a

27
Edition by Abásolo and Rodríguez (2005).
28
Carcedo (2008) arrives at an etymology *nei-b ho- continued in OIr. noíb
‘saint’, which is unwarranted in view of the inexistence of any evidence of
the change [u] > [v] or [b] and the consequent cases of hypercorrection
required for it to be true. Of course an original *nei-uo-, as in Gk. neiÒw
‘fallow land’, may have given rise to the attested form, but close cognates
and the specific derivation required remain in the dark.
29
It has a perfect match in the evolution of *gouÔ 2ro- > kor ‘curved’ and
*keuÔ 2ro- > sor ‘hole, cave’; the alternative reconstruction *neuro- (Beekes
2003: 165) has no firm comparative foothold. Cf. Martirosyan (2009: 373-
374, 739-740, with bibl.).
30
This could be the result of an ancient derivative *nouar-ko- or *nouar-
iko-, if this word underwent vowel syncope in the third syllable. *neu(V)ro-
would have yielded *núro-. There is a similar phonetic evolution in
*keuÔ 2ro-  *kouarno- > Lat. cauerna ‘cave’.

Volume 43, Number 1 & 2, Spring/Summer 2015


18 Blanca María Prósper

spurious laryngeal in an ani† root may be due to


contamination with the very ancient, arguably pan-IE
deadjectival, factitive present *neue-h2- in Hitt. neua%%-,
Lat. nouáre, Gk. neãv ‘to plow anew’, OHG. niuwon, and
possibly the Celtiberian FN nouantikum (gen. pl., K.9.3,
Numantia).
Since -ro- adjectives are often deverbative, it may be
argued that the old verbal form gave rise in a very early
stage of the protolanguage to a quasi-participial adjective
*neuÔ2-ro- ‘renewed’ > ‘young, fresh’ with suppression of
the thematic vowel and /e/ grade mechanically drawn
from its thematic base. This makes more sense than the
reconstruction of an endocentric, perhaps oppositive,
derivative in -ero- from the original adjective *neuo- itself,
which does not conform to the actual phonetics of the
attested forms. The laryngeal may in turn have been
metanalyzed and then spread to other adjectives; this
would account for *mad-Ô2-ro- (Skt. madirá- ‘intoxicated’
and Gk. madarÒw ‘wet, flaccid’) and other secondary
adjectives in -aro-. There is a river Novaria in northern Italy
after which the city of Novaria, today Novara was named
(Pliny, Nat. Hist. 3, 124; Ptolemy, Geogr. 3, 1, 29, etc.).
The fact that -eu- has not evolved into -ou- must be
explained in some way, either as a remarkable Celtic
archaism (cf. the place name Neviodunum in Carniola, the
lost epitaph to TEVTO / VITVLI / F(ilius) LEVCAN[-] in La
Rioja, etc.) or accepting that Nevara was an IE but non-
Celtic divinity. Some scholars, including Joseph himself,
have denied that a string *-eua- obeyed Joseph’s law (cf.
Zair 2012: 225ff.). Still, according to Schrijver (1995: 97-
100), *-eua- gave *-aua- by Joseph’s law, and this change
took place prior to -euo- > -ouo-. A good example is the set
of MIr. coar ‘hero?’, MW. caur, W. cawr (masc.) ‘giant;
hero’, Gaul. Cavarillos, the PN Kauarow, all of which come
from *kauaro- < *keuaro- < *keuÓ-ro- (cf. Skt. ßávíra-
‘powerful’).
A modern place name Navardún in northernmost
Saragossa (to the north of the Ebro river), from *Nauaro-
dúnum, conforms to the expected sound changes, but is
suspect of being the product of late infiltrations from Gaul,

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 19

like the Catalonian place names Verdú or Salardú. 31


Consequently, either the change is not Proto-Celtic, or the
divine name is not Celtic. Still, it may be the case that the
change was aborted by virtue of the analogical influence of
the simple adjective *neuo-. This must have happened in
the archaic ordinal-based name *neuano- ‘ninth’ continued
by the Gaulish PN NOVANVS, Celtib. dat. sing. fem. NVANE,
which did not evolve into *nauano- because the cardinal
*neuaN, from IE *neu÷ 'nine', had not undergone the
change (but see below). Under this assumption, *neuaro-
was preserved and then failed to undergo the change
*neuo- > *nouo-, which makes this divine name a
remarkable archaism if it is Celtic at all.32

31
The region name Navarra is first attested in the early Middle Ages.
Most scholars agree that it is ultimately related to nava from Celtic or IE
(but non-Celtic) *náuá ‘flat valley’, but the suffix remains unexplained.
Accordingly, it might be an ancient *nauar-iá with the same
resyllabification -r.i- > -r.ri- > -rr- as the northern ethnic names Susarri,
Gigurri, Seurri, etc. handed down by Pliny and in epigraphy (cf. Prósper
2008b).
32
Zair (2012: 225-43) has reviewed the rather messy and contradictory
evidence for and against preservation of the laryngeal in the context
CeiHC-, CeuHC- and concluded that laryngeal loss has taken place
preceding a stop and perhaps in the specific sequence CeiHn-. But Ceiar-,
Ceuar- at least seems assured. MEARVS CONTARI F (ilius) is attested in Vila
Real, MEARVS MEIDVENI in Cáceres and SVNVA MEARI in Viseu. Interestingly,
it is identical to the river Mearus, which was said to flow into the sea in La
Coruña (Callaecia Lucensis) by Pomponius Mela. This PN can only go
back to *meiÓ-ro- and the likely root is *meiH- ‘heranreifen, gedeihen’
(Pokorny 1959: 711-712, Rix et alii: 428), which has given a number of
adjectives meaning ‘soft, sweet’, such as Lat. mítis, OIr. mín ‘smooth’. This
testimony then seemingly belies the loss of laryngeals in this context
established from OIr. fíar ‘crooked, bent’ and MW. gwyr from *uei(H)-ro-.
Joseph’s law, by which *-eRa- yields *-aRa- in Celtic, is well attested in
Hispania: An interesting example is PALARVS (Castelo Branco, reading
unclear), PALARI (León, Vadinienses) and perhaps PALARIACVS (twice in
Venetia et Histria). The underlying formation is *k uelÔ 1-ro- ‘moving,
shifting’. A PN BALARVS in Ávila (Vettones), is rightly traced back to
*b helÓ-ro- ‘white’, but is hardly ‘Gaulish’ (cf. Zair 2012: 194). An apparent
exception TEMARI (Lugo, Callaecia Lucensis) is neatly explained if we
start from *temÓs-ro-, which yielded *temasro- and eventually *temáro- by
voicing and loss of the sibilant followed by compensatory lengthening (cf.
Skt. tami§rá-, Lat. tenebrae).

Volume 43, Number 1 & 2, Spring/Summer 2015


20 Blanca María Prósper

5. DVILLIS (Vaccaei). Goddesses of the netherworld


CL(audius) · LATTV/RVS · DVIL/LIS · V (otum) · S(olvit)
· L(ibens) · / M (erito) EX VI(su), Palencia.
ANINIVS (?) / AFRICANVS / CAERRI / AFRICANI F(ilius) /
33
DVILLIS / V (otum) / S(olvit) L(ibens) M (erito), Palencia.
The relationship of the divine name DVILLIS to OIr.
duille ‘leaf’ has been established on superficial grounds and
can be confidently ruled out.34 A more promising
etymology would start from the feminine *dubuí of the
adjective *dubu- ‘dark, black’ preserved in OIr. dub, OBret.,
W. du,35 which would yield *duuí. An alternative
simplification of the cluster would have been *dubí as in
the Gaulish river name Dubis, which in my view is due to
the influence of the masculine *dubu- (see below).
Somewhere down the line, this form was enlarged by a
suffix probably conveying a diminutive value, resulting in
*duuí-lá.
I believe the change -bu- > -uu- > -u- to be of Common
Celtic age. The sequence *dubuV-, where -V- is a middle or
open vowel evolves into *duuV- and then *douV- in all
contexts except before -i-, and, in our case, -í-.36 This

33
See Beltrán Lloris et alii (2007, with references).
34
It was first put forward by Albertos (1952: 55). Whether a southern
Italic gentilic Duilius is related is uncertain. Still another ex voto from
Cáceres reads VOTVM / FECIT LIBE(nter) / SELAIS · DVIL(lis? ) // DVLIVS /
IVLIVS (HEp 3 1993: 139) and is likely to be related, though the divine
name could be a singular DVILLAE. The dative in -<AIS > is neither Latin
nor Celtic, but the word may go back to an adjectival *sel-â/áuio- with a
latinized inflection.
35
Spread of the analogical zero-grade suffix *-u- is preferable to positing
a chain *d hub h-eu-ih 2 > *dubouí > *dubuuí preserving the original scheme
inherited by Gk. glukÊw, gluke›a, etc.
36
See Prósper (2007, Appendix), and for the required evolution -uu- >
-ou-, which is restricted to cases where no -i- immediately follows, McCone
(1996: 55). A good example of the simplification of the cluster is OIr. boí
‘was’, from CCelt. *boue, itself from *bubue and ultimately from the IE
perfect *b hu-b huH-e with the same reduplication as in O. fufans. Different
views in Jasanoff (1988): IE *b he-b huH-e > *b heb hue > *b heue > PCelt. *beue >
*boue, McCone (1991: 131-133): 3rd pers. sg. aorist *buuet > *bouet > *boue
> boí or a perfect, as reconstructed here, but with an evolution triggered
by dissimilation of labials; and Schumacher (2004: 250-254): *b hu-b huH-e
>> ICelt. *bube (for unclear reasons) > OIr. *buue (by dissimilation) >
*boue (‘weil das Altirische sonst kein *-uue- besaß’), which I find
unconvincing.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 21

provides the explanation of *duuí, on the one hand, and a


number of Celtic PNs on the other: Ogamic DOVAIDONA,
DOVATUCI, DOVAGNI, Hispano-Celtic DOVIDONA,
DOVIDERVS, DOVITERVS, DOVIDENA, DOVITENA (found
among the Cantabri Vadinienses, Lusitania and Celtiberia)
and Galatian Dobhdvn.37 The common origin of these
names is underpinned by the perfect match of the first
example, which continues an ancient compound *dubu-
aid-on- > *dou-aid-on- 'burning dark, dark fire'. From such
contexts a stem *dou- was metanalyzed; but a thematic
*dubu-o- has probably been at play here, too. The
contextual variant *douV- must have spread later to PNs
containing the string *dubui-, which need not have been
the case with the divine name DVILLIS. This would explain
the Gaulish PN DOVILLICCO (Belgica).
Therefore, Uhlich (1989) is in my view mistaken
when he states that Ogamic -o- instead of -u- in
DOVAIDONA, etc., is the product of a specifically Irish
infection or caused by the retention of a difficult cluster -
bu-, which he posits to account for a graphic discrepancy:
The outcome of CCelt. -bu- is <V> in Ogamic, but <B > in
the Latin alphabet (in the PNs DOBTVCI, DOBAGNI of Irish
ancestry attested in Wales). According to Uhlich, this
cluster finally yielded Irish [b] (witness the Irish PNs
Dubthach, Dubán, Dubaéd). By contrast, the lenited
outcome [b] of CCelt. -b- was uniformly rendered in
Ogamic as <B>, Latin <B >. Unfortunately, Uhlich never
takes into account the much earlier Hispanic and Galatian
names obviously belonging to the same set, where -o-
cannot be interpreted in primitive Irish terms and the
symbol <V> presupposes that -bu- had evolved into [u]. As I
have advanced, this scholar bases his idea on the later Irish
PNs Dubthach, etc. But, as their vocalism reveals, they must
have been remodelled on the synchronic adjective dub;
using them as evidence of an evolution -bu- > -bu- > -b- on
the grounds of their consonantism alone simply begs the

37
With regular fricativization of [u] > [b], like, for example, the PNs
Bhpolitanow < *uepo-litano- ‘broad faced’, Ordobetou, in my view from
*ordo-uexto- ‘hammer fighter’ (cf. the EN Ordovices).

Volume 43, Number 1 & 2, Spring/Summer 2015


22 Blanca María Prósper

question.38 Irish names of Wales like DOBAGNI, etc. have


been modified in their consonantism only, probably under
Brittonic influence, which has eliminated every trace of
[u], as in the river name Dubis. This is crucially backed up
by another PN DOBVNNI attested in Wales and an EN
Dobunní (SW. of Britain): Both forms go back to *dubu-n-
and consequently should never have -o- vocalism, were it
not because they have been redone in analogy to the
other PNs; in other words, because the variants *dou-,
*dub- brought about by sound change had been locally
subsumed in a compromise form dob-. Finally, a probably
Irish PN DOVITHI in Cornwall39 is reminiscent of the
Hispano-Celtic PNs DOVITERVS, DOVITENA and fatal for
Uhlich's distribution.
Coming back to DVILLIS, the doubling of -l- following
a long vowel can be phonetic or merely graphic. There are
some Gaulish parallels showing a suffix -illo- which cannot
be justified etymologically. There are some indications that
both DVILLIS and Gaul. -<ILL >- may be the product of an
original sequence -íl-. In fact, the sequence ‘high front
vowel + /l/’ is apt to undergo a third type of ‘inverse CL’:
In the words of Olive, Greenwood and Coleman (1993:
206-10), it is 'notoriously difficult to perceive the vowel-
lateral boundary in high, front vowel + clear /l/
sequences'. The Latin Paradebeispiel is the compound
tranquillus, which is taken to come from an adjective *kuílo-
identical to Old Norse hvíla ‘resting place’, Serbo-Cr. ç;li
‘rested’.40
Many of the Gaulish forms containing this suffix can
be traced back to the feminine stem of an athematic
adjective: SVADVILA (Aquitania) from *suáduí-lá ‘sweet’
beside the very common SVADVLLA and SVADVLLVS, beside
*suáduí-lá > *suádí-lá ( SVADILA in Aquincum, Pannonia,
CIL III, 10558) > *suádillá (*suádullá is the product of

38
In the same vein, Schumacher (2004: 250) posits for the cluster -bu- a
general evolution -bu- > -uu- > OIr. -b-. The only proof for the existence
of the intermediate stage -uu- is the PN DOVVINIAS, an unlikely input for
the alleged OIr. outcome -b-; no explanation is offered for the root vowel
/o/ in the onomastic material of Ogamic and Continental Celtic.
39
Cf. Sims-Williams (2003: 136, 183, nr. 1206).
40
See the dossier in Vine 2008, who reconstructs a phrase *trans ku ílim.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 23

crossing); MEDVILLVS from *medu-í as in, mutatis mutandis,


Skt. mádhví ‘sweet’, beside MEDIL( L)VS; MATVILLA
(Pannonia Superior), from *matu-í ‘good’ and ELVILLVS
(Belgica) from *(f)elu-í ‘many’. We may also reckon with
animal names supposed to have had a corresponding -í-
stem feminine at some time, such as MANDVILLA,
MANDVILLVS (Narbonensis, Transpadana, etc.) from
*mandu-í ‘she-donkey’, BANVILLVS (Aquitania) beside
BANILLVS, BANILLA, from *banu-í ‘sow’ (vs. masc. *banuo-).
Simplified forms showing no -u- may be explained as based
on thematic forms after elimination of the original -u-stem,
or as reflective of regular contextual deletion.41 The non-
inherited Gaulish suffix -illo- was probably metanalyzed
from -í-stems, where etymological -lo- may have been more
frequently used given its diminutive sense (with the
possible contribution of -i- and -io-stems). Reanalysis was
probably favored by a tendency to eliminate or simplify the
unproductive morphological opposition -u-s: -(u)u-í: -u. -illo-
may have been selected among other possibilities in view
of its comparative transparency and then become
independent and productive, so there is really no way of
recovering the formational details of every particular name
after the process was triggered.42

6. AIRONI (Olcades). Superficial matches that cannot be


taken seriously
DEO A/IRONI / FECIT FA/MILIA OC/VLE(n) S(is)
VSE(tana?) / C(aius) TITINIV [ S] / CRISPINV [ S ], Uclés,
Cuenca, CIL II, 5888.
The divine name AIRONI can be classified as Indo-
European since we know for a fact that the area was
inhabited by the Olcades, whose PNs are unmistakably
Celtic. We may add to this isolated attestation a graffito
from the cave La Griega in Pedraza (Segovia) which reads

41
The synchronic pronunciation may have been -uuillo- in 'Sievers'
contexts, witness MANDVBILI (Dijon, CIL XIII, 5532 if not a compound) vs.
MANDILONIS (Brixia, CIL V, 5001).
42
Its origin is different from that of Latin -illo-, which as is well known is
the result of samprasárana in sequences including the Italic diminutive
suffix -elo-, as in *ag un-elo- > auillus (vs. the simple form agnus) and *bak-elo-
> baculum, but bacillum with a double suffix -elo-.

Volume 43, Number 1 & 2, Spring/Summer 2015


24 Blanca María Prósper

GAIVS [-5-] AIRON[--] VALER[--] (Mayer and Abásolo 1997,


nr. 97).43
AIRONI probably goes back to the IE heteroclitic
formation *h2ei®-, *h2ei-en- ‘day, morning’, attested in Av.
aiiar/n- (cf. Wodtko et aliae 2008: 258-259) and a number
of derivative formations. Greek and Germanic are
traditionally held to have inherited a locative *h2ei-er-i,
resp. in the neuter compound êriston ‘breakfast’ and in
Goth. air, Old Norse ár 'early'. A secondarily thematic
*h2eiero- is attested in Av. aiiara- and designates the
divinities of the day, and note fraiiara- and uzaiiara- 'early,
morning/afternoon'. The noun in question is not attested
in Celtic.
A divinity of the same etymology is attested among
the Zoelae (a branch of the Astures whose capital was in
Castro de Avelãs, close to Bragança, the ancient Brigantia
in Portugal) who worshipped a DEO AERNO preserved in
four altars, three of which were found in precisely this area
(cf. Garcia 1991: 281-282). This form can be traced back to
a trivial thematic adjective *h2eieri-no- or *h2eier-no- built on
the locative of the noun for ‘day’ (cf. Prósper 2002: 283-
286), but the story may have been more complex, as we
will see below. As Redentor has remarked (2006), Juppiter
is the most frequently mentioned divinity in this area
beside AERNO, which is consistent with the idea that an
indigenous divinity of daylight was reinterpreted in a
Roman guise. Whether this is a Celtic divinity (both in the
linguistic and the cultural sense) or not remains open.
It is attractive to derive both AERNO and AIRONI from a
single source: An ancient athematic nasal stem which is

43
In spite of what the edition asserts, a personal name Airo, -onis does not
exist. Due to the poorly preserved state of most of the text, the structure
remains unclear. In this cave there are dedications to other Celtic
divinities such as DEVAE and NEMEDO. Incidentally, some considerations
about Celtiberian phonetics make it unlikely that NEMEDO, attested twice
in these graffiti, goes back to *nemeto- ‘sacred place, sanctuary’ as
currently assumed (cf. Marco Simón 1993). Even allowing for sanctuaries
themselves being adored, which is debatable, there is no voicing of
intervocalic voiceless consonants in the whole area. Additionally, there is
an equivalent dedication to NEMEDO ASEDDIAGO in Ujo (Asturias), but the
Astures held a fortress named Nemetobriga according to It. Ant. 428, 6 and
Ptolemy, Geogr. 2, 3, 21, which makes the contradiction still more blatant.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 25

not likely to have been directly formed on the nom./ac.


sing. *h2ei-®, both because the natural option would be the
oblique stem and because the vocalism of the suffix would
probably be reflected as -ar. In other words, it is not likely
that the outcome of an ancient *h2ei®-h3ó(n) or a more
recent *aiVr-ó(n) would have given this final form.
Alternatively, we may depart from the locative form. An
endingless locative *h2 ei-er, which seemingly forms the
base of Av. aiiara-, may have been hypostasized as an
exocentric nasal stem *ai-er-ó(n) ‘possessing the daylight’
which is amazingly reminiscent of the Iranian divine
name.44 The apparent retention of a diphthong -ai- in the
form AIRONI (no <AE > or <E > is attested) indicates that we
are not dealing with an ancient CCelt. diphthong /ai/.
<AI> is probably reflective of an early contraction *aier- >
*aiir- (> *air-).45
The locative of IE *h2ei-® is universally held to have
been *h2 ei-er-i. This idea is exclusively based on Gk. êriston
‘breakfast’, which allegedly contains *ári from *h 2eieri and
a past participle *h1 d-tó- ‘eaten’ (recently Beekes 2010:
131-132). This idea has something to recommend itself
only as long as a very ancient compound has given PGreek
*árísto-, with /i:/ as the regular outcome of a non-final
sequence -iH-.46 Some authors favor a compound of *ári +
estó- with a vocalized laryngeal, which is unlikely to have
yielded the attested result. There may be a solution
matching the extant evidence:m *h1 d-tó- should have
given Gk. †fistÒ- if we trust the testimony of the imperative
‡syi from *h1s-dhi ‘be’ (identical to Av. zdí; see Rix 1992:
263).
There is an increasing amount of evidence showing
that the regular result of /h1/ preceding a consonant

44
Cf. Nussbaum (1986: 235-238), and loads of recent literature in his
footsteps.
45
Cf. the accepted CCelt. change -eie- > -í-. The strings -áie- and -áié- seem
to yield Hispano-Celtic -á- (cf. ‘Vasconian’ A.39 bentian from a
univerbation of locative + postposition *penxtiái-eni and Celtiberian K.1.1
asekati from a subjuntive *ad-spek-á-(i)-é-ti). Cf. Prósper (2013b: 7).
46
Note that acceptance of the ‘Francis-Normier rule’ would not
contradict this result, since it only asserts that a word medial sequence
-i/uh 2/ 3- yields respectively the diphthongs -i/u-á/ó-.

Volume 43, Number 1 & 2, Spring/Summer 2015


26 Blanca María Prósper

cluster was (h)i-. De Vaan (2009) has argued that Gk. ·ppow
‘horse’ goes back to *h1kuos, with the vocalism
corresponding to the original genitive of the athematic
*h1eku-, preserved in Anatolian only, as opposed to the rest
of IE, which has regularised this form as *h1ekuos; Gk. fixyw
‘fish’ equally goes back to *h1 dh ghuH- (*dh ghuH- would yield
†xyw).
All this partly draws on ideas by Vine (1999), who has
made an ingenious case for the development of a Proto-
Greek segment -i- in certain environments, e. g. fipnÒw
‘oven’, from *s.pnó-. In fn. 42 he reckons with a preform
*H.pnó- as an alternative, but dismisses it on the grounds
that the aspirated variants of fipnÒw would remain
unexplained by this account. 47 Accordingly, all these
recent works militate in favor of the evolution proposed
above for *h1 dtó- > *(h)istó-. This word might be, apart from
structural considerations, one of the very few forms
illustrating the existence of an initial laryngeal in the root
*h1ed-, beside nhsteÊv ‘to fast’, n∞stiw 'fasting, jejune' and
A
Toch. nätsw- ‘to starve’, from a privative compound *÷-h1 d-
tu-/-teu-io-.48
In sum, êriston may go back to early or pre-Greek
*aier + istó-. If this is correct, the reconstruction of a
locative *h2 eieri in other IE branches loses ground: The
preform *aiiri posited to account for the Germanic forms49

47
Recently, Bozzone and Felisari (2012) point out in Vine’s footsteps that
the actual outcome of *h 1 CC- via *h 1e 2 CC- is *hiCC-, with an aspirated
Anlaut and with vowel epenthesis, which took place prior to laryngeal
vocalization; this laryngeal leaves an aspiration that is retained except
where Grassmann’s law comes into play (which in their view eliminates a
hypothetical *h 2/ 3 .pnó- for fipnÒw, where the required laryngeal is not
allowed by Hitt. %appina-). Their overarching conclusion is that both PIE
and Pre-Greek *h 1 i- yield *hi-, as opposed to *h 2/ 3 i- > *i-.
48
See Adams (1999: 493). The composition vowel in the Greek nouns
deipnhstÒ/Êw, dorphstÒn ‘lunch, dinner time’ is unclear, but these forms
may have been refashioned from regular *-o-h 1 d-to- > -vstÒ- after the type
»mhstÆw and »mhstÆr ‘eating raw meat’, Skt. ám-ád-, from *h 1éd-.
49
See Orel (2003: 9) and Kroonen (2013: 12), who take Gk. ±°riow ‘early’
to be its direct cognate, which is probably wrong. Kroonen additionally
takes Old Norse ádr 'early' from *h 2ei-er-is, but this is probably an entirely
different word.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 27

is not more plausible than *aiir, with reduction of IE /e/


after -i-, and only supported by the Greek compound.
Goth. air and ON. ár ‘early’ consequently may go back to
IE *aier. As regards OE. ǣr, ON. ær and OHG. OS. ér
‘earlier, before’, their meaning indicates they continue
the PGerm. comparative form *airiz in Goth. airis; in spite
of the comparative meaning 'earlier' of the western forms,
these are usually referred to as if all of them could go back
both to *aieri and to *aieris, which is unwarranted but
suitably fills all the formal slots.
As we have seen, there is no compelling evidence for
the reconstruction of a locative of this word other than
*h2eier (possibly from *h2ier with early paradigm levelling),
which was distinguished from the nom.-ac. merely by ablaut
and probably accent. The process giving rise to the nasal
stem is therefore one of external derivation, by which
post-IE *aier is enlarged by one of the two basic
holokinetic nasal suffixes, or alternatively a secondary,
unattested locative *aier-en is hypostasized as *aier-ón (cf.
*gh eim-én  *gh eim-òn). On a slightly different account, the
word on which this divine name is based would be the same
as gave rise to Av. aiiara-, which is just a thematic derivative
from the (endingless) locative just like Gk. ˆneirow from
*h3(V)neri-o- or m°sow from *medhi-o-. An individualizing
nasal stem was later built to this primitive thematic
formation.
Other formations which used to be said to contain
this locative, like Gk. ∑ri, ±ri-g°neia, and its derivative ±°riow
‘early’, are currently traced back to *h2eus-er-i, the locative
of a hysterokinetic formation *h2eus-ér, in turn built to a
still more primitive locative *h2eus-er 'in the morning'. All
this points to the -i locative being often restricted to
secondary formations. In our case, a reconstruction *aieri
has been passed down from the first etymological studies to
the present not only because of the aforementioned
adjectives, but also because laryngeal phonemes were still
largely unknown, and consequently a past part. *dtó- or
*edtó- could be reconstructed for a root *ed-.
AIRONI being a comparatively ancient formation, its
nasal suffix is not certain to be the product of the
adaptation of -ún- to the Latin declensional patterns as

Volume 43, Number 1 & 2, Spring/Summer 2015


28 Blanca María Prósper

usual for personal names, but may be reflective of an older


construction: the expected oblique stem of this
holokinetic noun would be *aier-n-. Accordingly, it is
conceivable that it gave rise to the thematic *aier-n-o- very
early, and that this form is preserved in AERNO with loss of
medial *-i-. In a different branch, *aier-n- was regularized as
*aier-on- and then its medial syllable underwent vowel
raising, yielding *aiir-ón-. Whether the nasal addition is
Celtic or earlier is impossible to decide, and accent
differences are possibly involved. Summing up:
Loc. sing. as adverbial: *h2ier >> *h2eier > PGerm. *aiir
> *air
Loc. sing. in compounds: *h2ier >> *h2eier (> Pre-Gk.
*aier) + past part. *h1 dtó- (> Pre-Gk. *(h)istó-) = *aiér-isto- >
êriston
By external exocentric derivation *h2eier  *h2eier-o- >
Av. aiiara-, beside fraiiara, uzaiiara- (hypostasized from
prepositional *pro-aier?)50 and perhaps PGerm. *airaz,
*airuz ‘messenger’51
By external exocentric derivation *h2 eier-o-  *h 2eier-
ó, h2 eier-n-ós (>> Hispano-Celtic *aier-ú, *aier-on-)
By further thematization  ?Hispano-Celtic or
Lusitanian *aiér-n-o-
Recent works relate this divinity to the waterfall called
Pozo Airón in Pereña (Salamanca), the well equally called
Airón in La Almarcha (Cuenca), etc., and are unfortunately
becoming fashionable even in academic circles (cf. García
Alonso 2010, Abascal 2011). The idea goes back, as far as I
know, to the naïve but understandable identification
posited one century ago by Quintero Atauri (1913), the
amateur archaeologist who actually uncovered the altar.
These hypotheses are not only far-fetched, but next to
impossible, and simply ignore the basics about the
historical evolution of the Spanish language: they demand
the preposterous premise that a pre-Roman diphthong [ai]

50
In a similar vein, Dunkel (2007: 56 fn. 15) posits a (later refashioned)
hypostatic *pro-aien. Note that the expected locative *aien(-i) built on the
oblique stem remains unattested.
51
This form cannot be derived from *airi (cf. Orel 2003: 9), but perhaps
goes back to PGerm. *air or to *aieros > *ajiraz > *airaz.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 29

for some reason failed to evolve into [aI] > [æ], > VLat.
[e] and finally Sp. [ie] or [e] (according as the vowel was
accented or not) in this particular word everywhere in the
Iberian Peninsula.52 An IE root *air- cannot be posited on
phonotactic grounds, since it contravenes Meillet’s rule
prohibiting contiguous sonorants in coda position. In spite
of the sizable number of historical (Celtic) and prehistoric
(conventionally labelled Alteuropäisch) river names attested
all over Europe, there is no reconstructed stem of this form
which has an aquatic meaning. The river whose spring is
close to the area is called Bedija, a likely derivative *Bedisiá
from Celtic *bedo- ‘ditch, grave, basin’.
As to the proposed etymologies, García Alonso (2010)
reconstructs Celtic *ariú ‘lord’, which requires a Celtiberian
metathesis -ri- > -ir-, for which there has been found not a
single reliable example thus far (in point of fact, all the
extant examples point to -ri- > -r.ri-). Why pitfalls or caves
should be called ‘lords’ remains an open question. In fact,
the existing well called Pozo Airón in Cuenca and the place
where the inscription was found are separated by around
70 kms. Close to this place there is a cave known as ‘cueva
del Aire’, which simply refers to the buzz or whistling one
can hear from the entrance and is of course unrelated to
the divine name. Airón is attested more than one hundred
times and applies to places and roads. Its preservation in, e.
g., Pt. Airão and a river L’Airon (France) militates against
the proposed identification. Additionally, Airón in Cuenca
has a doublet Lairón which puts us on the right track: We
are dealing with an old Spanish appellative that has fallen
into disuse, and this is the reason why the definite article
el has been metanalyzed as part of the noun. There is

52
According to García Alonso (2010) the anomalous evolution is due to
paretymology of Sp. aire, which is a notable coincidence taking into
account the sizable number of names Airón. But that regular †Erón should
have been ousted by Airón everywhere is untenable, even allowing for a
number of secondary place names. His observation that the phonetics of
aire equally show lack of monophthongization is beside the point, since,
to begin with, Lat. aer is of Greek origin (éÆr) and contains /a:/ and
consequently a hiatus, and secondly, Sp. aire is not a patrimonial word that
fails to evolve for some reason, but was adopted from written Latin at a
late date along with many others, as both the secondary diphthong ai- and
the preservation of final -e indicate.

Volume 43, Number 1 & 2, Spring/Summer 2015


30 Blanca María Prósper

another pit called Los Airones in Toledo, which leaves no


room for doubt as to its appellative nature. Airón may be
the sporadically employed augmentative of aire, which used
to mean ‘whirlwind, whirlpool’ when related to water;
whether some of these instances is the product of
paretymology, and consequently conceals an older
indigenous name, it is no longer possible for us to
ascertain. This should be the end of the identification of
AIRONI with Sp. Airón.

7. Reconstructing the Celtic paradigm of IE *terh1/2- with


the help of onomastics
Rix et alii (2001) distinguish two different but nearly
homonymous and synonymous roots: *terh1- ‘durchbohren,
reiben’ (632-633) and *terh2- ‘durchkommen, ûberqueren’
(633). In most languages and phonetic contexts, the
results of these roots are very difficult to distinguish. The
Celtic evidence is sparse and can only be interpreted
according as the form belongs to a transitive or an
intransitive verb. Thus, OIr. tráth ‘period of time’, MW.
trawd ‘journey’ go back to *t®h2-tu-. On the other hand,
MW. trawd, Bret. treut ‘weak’ are traced back to *t®h1-tó- and
identified with Gk. trhtÒw ‘bored through’ (see Zair 2012:
82 with references). A PN in the gen. sing. CONTRATAI in
Cáceres goes back to *kon-trát-áio- or a syncopated *kondo-
trát-áio- ‘mentally weak?’ and is a possible cognate of MW.
trawd. According to Schumacher (2004: 620), *terh2- is also
continued in the Celtic present formation *tar-io-
‘sprunghaft hervorkommen, bersten’, attested only in
Brittonic (MW. tardu, OBret. tardom), which in turn goes
back to IE *t®h2-ié/ó-. A participle *t®h2-÷t- ‘crossing’ is only
preserved in two close cognates: Latin trans ‘through’ and
OIr. trá ‘then’. The unexplained potter’s name VATRAVNVS
in Belgium may be analyzed as a compound *u(f)o-trámno-,
in turn from a middle aorist participle *t®h1/2-mno- 53 .
Nussbaum (2010) takes recourse to a similar participle *trH-
mno- to explain the Greek adjective tranÆw ‘clear’.
53
CCelt. *tramno- may alternatively be a regular derivative built on the
genitive *tra-mn-ós of an agent noun *tera-mon- (> *tara-mon-), but this
formation is everywhere based on a laryngealless variant of the root, as
we shall see below. Gk. terãmvn ‘weak’ is not certain to be related.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 31

According to him, this example shows that the


posconsonantic cluster -mn- is simplified even when the
preceding consonant was a laryngeal. This idea may need
refining, in view of the Celtic names in *-amno-, of which
this is just an instance. Since the disappearance of
laryngeal segments is currently understood as a
comparatively late phenomenon, it may be the case that
the Proto-Greek structure evolved as *t®H-mno- > *t®Hno- >
*tráno-, as opposed to CCelt., where *t®H-mno- would yield
*trámno-.54 The Greek form occurs in composition with
prefixes, as in the late forms §ntranÆw, per¤tranow ‘very
clear’; the privative étranÆw ‘unclear’ may have a very
similar meaning to that of VATRAVNVS.
The PN in the gen. sing. TRANTIONIS 55 is perhaps the
only onomastic remnant of an active participle *t®h1/2-÷t-
‘crossing’ preserved in two close cognates: the Latin
preposition trans, the Old Irish adverb trá ‘then,
therefore’, and, with preservation of final -s, the adverb
and adjective MW. traws ‘across; violent, implacable’
(Schumacher 2012, Zair 2012: 179); the actual
phonotactics are unclear, however, and *trÔ1/2-nt- cannot
be ruled out. Celtic *trâ/ánt- consequently looks like the
active counterpart of *tramno-.
Recently, a most intriguing text has come to light in
El Salobral, La Vera (Cáceres, Lusitania Emeritensis).
According to the editors (Gómez-Pantoja, Madruga and
González Cordero 2013: 285-287), it reads (as the
photograph seems to confirm, at least concerning the first
three lines):
D(eo) ET D(eae) / VLISO IGO / TIIRIIVN / VL I §O /
NIAII / AQVILVS / MODESTI / F(ilius) V(otum) S(olvit)
L(ibens) M (erito)
After considering several segmentations, the editors
posit a dedication to the paredric divinities VLISO and

54
Alternatively, *trÓ-mno- would yield *trâmno-. Note that Celtib. launi
militates against early simplification since it comes from *uÒh 1-mn-ih 2
(see the dossier in Prósper 2014a).
55
BATO trantionis F (ilius) / ARAVISCVS ANN(orum) L / H(ic) S (itus) E(st)
FIRMVS H(ic) S (itus) E(st) / MOGITMARVS T(itulum) M(emoriae) P (osuit), CIL
III, 3325 (Pannonia Inferior, Also-Szent-Ivany, Hungary). Delamarre
(2004: 281) adds a PN TRANTO (a mistake for the Pannonian name?).

Volume 43, Number 1 & 2, Spring/Summer 2015


32 Blanca María Prósper

VLISONIAII and reasonably isolate an epithet IGOTEREVN (o)


determining VLISO; this form is said to be comparable to a
number of Gaulish compounds whose first term is ico-.56
Some of these contain clear warlike nuances, like DEAE
ICOVELLAVNAE (Belgica). Although it is in principle
possible that -tereuno- is directly related to *uo-trauno-, the
vocalism poses questions. 57
Stifter (2006) has shown that the result of -nm- in
Celtiberian is -lm-, with a characteristic tendency to merge
the lenited result of Celtic /b/ and /m/.58 Elsewhere
(Prósper, 2014b) I have contended that a similar
dissimilation equally applies to -mn- in contexts where it
cannot regularly yield -un-, as in the string -rmn-: I have
proposed that Celtib. terberez in the much debated Bronze
‘Res’ reflects *ter-mn-ed, the ablative of a Celtic holokinetic
masculine noun *ter-mú, and that it had probably gone
through the stages [terbned] 59 and [terbred],60 an
evolution that resembles, mutatis mutandis, that of Latin
hominem > omne > omre > Sp. hombre.
This noun would be ultimately identical to Lat. termó
‘finishing post in a race’, Gk. t°rmvn ‘boundary’, the divine
name TERMVNIBVS AVG(ustis) (Noricum, CIL III, 5036,
probably a Celtic form with spread of -ú from the nom.
sing.), the Venetic divine name termonios deivos (acc. pl.

56
Bear in mind that voicing is an areal feature affecting the best part of
western Hispania.
57
Unless we accept it goes back to *treu-(m)no- from *treh 1u- ‘to feed,
rear’, Rix et alii 2001: 647 or perhaps *treud- ‘to push’ in MW. cy-thrudd-
‘to bother, disturb’, MIr. trotaid ‘fight’, Rix et alii (2001: 651). This
possibility has the obvious disadvantage that the diphthong /eu/ would be
preserved contrary to expectations. Nussbaum (1997) traces Gk. t°ru
‘weak’ back to a noun *terh 2-u-, with an extension of the root which is
unlikely to form the base of the Hispanic *tereuno-, in spite of its attractive
similarity to Gk. terÊnhw ‘decrepit’, Skt. taruna- ‘soft’.
58
Stifter admirably clears up the PN Melmu < *men-mú(n) vs. Gk. M°mnvn,
and albana < *anmana ‘names’ in K.1.3 (Botorrita), as opposed to Gaul.
ANVANA in Larzac, but SVANMANV ‘good named’ (tabooistic for ‘seven’) in
Rezé.
59
Cf. in a lexical boundary Gaulish COBNERTVS from *kom-nerto-.
60
I use the Greek alphabet for the transcription of Celtic lenited
consonants.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 33

masc.), Palaeo-Italic (Oenotrian) termani[-61 and Skt. su-


tarman- ‘crossing well’, besides the secondary thematic
forms Lat. terminus, U. termnom, from *termno-. 62 Although
this form can probably be traced back to *terh2-, it never
shows any trace of a laryngeal, perhaps because both
nominative and accusative of the agent noun would
undergo the Saussure effect (that is to say, laryngeal loss
due to the suffix having the form -mon-) or because in the
oblique cases the laryngeal precedes two consecutive
nasals.
In order to explain *tereun-, I propose to start: a) from
a thematic *termno- (cf. Lat. terminus, U. termnom) or b)
from a regularized gen. sing. *ter-mn-ós of *termon- (with
secondary full grade as in the abl. sing. *ter-mn-ed above,
instead of the expected *t®(h2)-mn-ós). 63 This string
evolved into *termno- and underwent dialectal anaptyxis by
way of a copy-vowel (Oscan teremnatust, teremenniú, cf.
Untermann 2000: 745) yielding *teremno- and then by
dissimilation *tereuno-, directly represented by
64
TIIRIIVN(---).
According to an alternative scenario, *termno- was apt
to be pronounced as *terbno-, and then [b] in the
resulting *terebno- was vocalized to [u] in coda position in

61
Cf. the exhaustive analysis of this form in Lazzarini and Poccetti (2001:
149-151): Since the ending is erased, many options are open to
consideration: it may be a -io-derivative like the Venetic form. The
vocalism of the medial syllable may be the result of spread of -man from
-m÷ to the rest of the declension, or a well attested phonetic hesitation
a/o, and then from -mon-.
62
RV VIII, 42: sutarmánam adhi návam ruhema ‘we would ascend on the
well-crossing ship’, from the masc. acc. sing. *ter(H)-mon-M. The account of
the Italic forms in De Vaan (2008: 615) is imprecise and somewhat
arbitrary. Explaining away termó as a Greek loanword poses more
problems than it solves.
63
If the epithet is abbreviated, the whole form may have been something
like *iko-termn-iko-, but, in contrast to most western Hispanic structures of
divine name + epithet, it does not have the appearance of a place name or
area protected by the divinity. See below for an alternative.
64
An etymological relatedness to epigraphic Latin TEREBVNIVS,
TEREBONIVS (Italy) can be dismissed. *treb- would not have undergone
anaptyxis in the languages of the area. On these PNs see Adams (2007:
75).

Volume 43, Number 1 & 2, Spring/Summer 2015


34 Blanca María Prósper

order to optimize the syllable structure -C.R- (which is


unstable because the onset of the third syllable is weaker
than the coda of the second), yielding *tereuno-. It is the
same evolution as in Latin tabula > Early Romance tab.la >
Catalan taula (as opposed to resyllabification in Sp. tabla,
Fr. table and preservation or anaptyxis in It. tavola). In
other words, while Celtib. terberez (ideally Proto-Celtib.
/termned/) has apparently resolved the phonotactic
difficulties by way of dissimilation, Western Hispano-Celtic
has done it by the alternative way of vowel insertion, thus
preventing increasing consonant differences within the
same paradigm and the rise of new morphonological rules.
The Galatian place name Iko-tar¤on (near Ankara) may
contain a derivative of *t®h2ó- ‘crossing, traversing’ and
have a similar structure and meaning, and *î/íko- (or
*(f)î/íko-), whose etymology remains obscure, may have
designated some kind of weapon by whose power the
divinity ‘comes across’, which is a metaphor for ‘rules,
obtains victory’, etc. or, if the meaning is causative, ‘helps
(men) across’.
Accordingly, it is possible that the compounds Skt. su-
tarman- and Celtic *?iko-termon- are the continuants of the
primitive procedure of creating compounded possessive
derivatives in -mon- by internal derivation from verbal
nouns in -m÷, in this case *ter-m÷ (Lat. termen, Gk. t°rma
‘boundary’, Oscan teremenniú, etc.); see Nussbaum (1986:
120) for other examples of the same derivational chain.
Contrary to most opinions relating the divine names from
*termó exclusively with boundaries as if their meanings
were directly derived from the Greek and Latin nouns,
these divinities may have had a function of helping
persons ‘come across’, whether this is understood as
leading the souls in their last voyage or protecting
travellers and paths.65 There is some evidence of a Latin
divinity Termó in the Livian Periochae 13, 19: ‘Termonis et
Iuventae arae moveri non potuerunt’.66

65
The Rig-Vedic form is essentially ambiguous in this respect.
66
An inscription on instrumentum domesticum from Durocornovium,
Britannia reads VENI TERMO MANE (RIB 2-3, 2422), which, if interpreted as
‘come, Termo, stay’ might be ultimately related. It has been read as MANE

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 35

Finally, one may envisage the interesting possibility


that the form is not to be read as IGO/TEREVN, but as
TEREVN/IGO. Not only are divine epithets all over western
Hispania endowed with a velar suffix of the form -(a)iko-,
but this suffix is also usually expressed, not abbreviated.
The text in the second line reading VLISO IGO seems to
have been accompanied with difficulties between the first
line reading D ET D and the third line containing TIIRIIVN,
and the upper borders of TIIRIIVN are very close to the
upper line. This would make sense if the entire line was
added with hindsight to match the rest of the text when it
dawned on the scribe that there was no room left for the
suffix in the fourth line, or if the scribe started writing
VLISO too close to the left and when he saw that he could
not make room for TIIRIIVNIGO in the next line, margine
urgente, he used the available space in the upper line. VLISO
TIIRIIVNIGO would make acceptable sense and be in line
with the more usual, transparent structure of these
dedications.67 But, if this were the case, the epithet would
possibly be derived from a place name (called either
*Termú ‘the boundary’ as in the ‘bronze Res’ or *Termno-)
instead of alluding to a trait of the divinity, and his
relationship with the Italic divinities as such would be
destroyed.
Note that a hitherto neglected but possibly similar
structure underlies the Gaulish divine name Sucellus, well

TERMO VENI, an invitation ‘come, T(h)ermo(don), in the morning’


(Birley 1980: 139), which I find unwarranted. There was a consul called
M. Minucius Termo, but the cognomen is never attested in epigraphy.
67
The suffix of the epithet is frequently relegated to the next line all
over western Hispania: MALVN/AICO, ENDOVELL/ICO, SEGIDI/AECO,
TANGINI/CIAECO, VEIGEBR/EAEGO, VAELI/CO, PARAMA/ECO, VORDE/AECO,
ROVD/AECO, D/VRBED/ICO, BORMANI/CO, VDVNN/AECO, VELI/CO,
TOIR/AECO, OENAE/GO, SENA/ICO. This may be due to a deeply
entrenched habit in the organization of the text that was sensitive to the
importance of the place protected by the divinity and accorded it a
relevant position, occupying the whole line when the altar was narrow.
Accordingly, if something went wrong after the ordinatio of the text,
TIIRIIVN would have occupied all the available space as expected. If the
second VLISO was already written, too, there was no room for the suffix.
We could even read the original text as a symmetrical VLISO /
TIIRIIVN(igo) / VLISO / NIAII (go), where only the first suffix was restored;
but d et d would not be understandable.

Volume 43, Number 1 & 2, Spring/Summer 2015


36 Blanca María Prósper

attested in Germania, Belgica, Narbonensis, etc. Although


it is usually taken to be a karmadháraya 'good striker', or a
karmadháraya-bahuvríhí ‘having a good hammer’ on
account of its iconography, since there are statues in which
he is represented bearing a hammer or mallet, he is also
depicted holding a pot (see Delamarre 2003: 113 with
bibl., etc.), it is conceivable that this is an older formation
in which cell- is of verbal origin, as in Skt. su-tarman-. In
lieu of an IE root *kelH- ‘to smite’ I would posit *kel- ‘to
protect’. *su-kel-mó(n) is the god ‘having a good protection
or *kel-m÷' > agentive 'protecting well, providing good
protection’ and the expected thematic result built on the
oblique stem is *su-kel-mn-o- with simplification and
assimilation of the sonorant cluster and a secondary full
grade of the root. 68 The advantage of this proposal is that
there are many compounds from this root in different
Indo-European branches, showing its derivatives were
often applied to the protection given by divinities, as in
the Luwian thematized PNs Kubanta-zalma, Tar%unta-
zalma, etc. and also in the Cilician priest names Iazarmaw,
Trokozarmaw (note the alternative thematization with loss
of the segment -n-); these names are often described as
possessive compounds containing the action noun *kel-m÷,
which is only true from a semantic point of view; in fact, the
inflection type underlying these compounds must have
been the masculine posessive *-kel-mó(n), not the neuter
from which it is derived. The athematic formation is
preserved in the Skt. PNs Soma-ßarman-, Prajápati-ßarman-
or in OHG. Ans-helm (from a thematic *helmaz). PNs whose
first term is a divine name must have been understood as
conveying a passive meaning 'having, obtaining the
protection of'. Crucially, the athematic formation
presupposed by Sucellus is identical to the masculine PN Su-
ßarman-, well attested in Hindu mythology and also a name
of a class of gods in the plural (Márkan∂eya-Purána).
If the reading of the edition is right, the divine name
VLISO stands a good chance of being the Hispano-Celtic
outcome of a past part. *uÒd-tó- ‘sought after, feasted,

68
Note that if OIr. soichell 'generosity, kindness' is a cognate of Sucellus, the
traditional etymology is untenable.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 37

honored’ it would be the only cognate of OIr. fled, etc.


‘feast’ < *uÒd-eh2, Gaul. VLIDORIX, Gk. ¶ldomai ‘desire’.69 All
this is suggestive of a Celtic verb *ueld- having existed in
Proto-Celtic. Adopting an idea by Michael Weiss, Pike
(2011: 106-107) has recently pointed out that ¶ldomai may
not contain a formant -d- alternating with -h1-, but was
actually added to *uelh1- 'to want, desire'. If this root
showed Narten properties, the long grade in a variant
*uélh1 d- would undergo shortening after laryngeal loss
(which in this context would be comparable to the
Saussure effect causing loss in the neighbourhood of /o/).
Conversely, Lat. volup ‘agreeably’, voluptas ‘pleasure’ shows
vocalization of the laryngeal as expected in a sequence
*uelÔ1p-. To be sure, this is an ingenious construction, but
if VLISO is related, Celtic must have built a neo-ani† past
participle to the synchronic *ueld-, since in principle we
would expect CCelt. *ulásso-; the same zero-grade
underlies OIr. fled, etc.
Accordingly, the divine name SADV / VLAD/V on a
stone found in 2003 during the excavations of the
indigenous castellum of San Cibrán de Las (Orense,
Callaecia) is probably unrelated. VLADV may be the dative
form of Celtic *ulato- 'prince, ruling' (in turn from *ulH-tó-
'powerful, ruling') with voicing of the intervocalic stop.70 It
might be a phrase *sádúi ulatúi but also a compound from,
e. g., *suádu-ulato- or *sádo-ulato- 'mild, benevolent ruler',
and, if so, comparable to Gaul. ATEVLATOS 'grand lord' on
coins. Interestingly, the last <V > in the first line is only
half the size of the rest of the letters, and it gives the
reader the impression that this line is shifted to the right
in comparison with the second, so that, if we dispense with
it, the text is centered. If <V > is secondary, the original
text would reflect *s(u)ád(u)-ulato- with a vocalic outcome
of u- in this context, and the addition is intended to
restore the -u-stem morphology of the first term.71

69
Reduction of a primitive *uelis- is conceivable but thus far only two
instances of VELISA (in Etruria and Alpes Maritimae) have been recorded.
70
Cf. the Gaulish PNs VLATOS, VLATTIVS, VLATVNVS.
71
The editors' comments (Álvarez González et alii 2004: 239) are
confused and inconclusive: They accept the possibility of a 'single word',
and mention an inexistent IE root *sed- 'peace' (this meaning is secondary

Volume 43, Number 1 & 2, Spring/Summer 2015


38 Blanca María Prósper

A number of Gaulish PNs are traced back to IE *t®h2ó-


‘crossing, traversing’. Many of them are compounds, such
as BROGITAROS, DVTARIVS, Dhiotarow, etc. (see Delamarre
2003: 291). This form corresponds formally to Skt. turá-
‘rapid, fast, impetuous, strong’. To this set we may add the
father’s name in an inscription reading MEARVS CONTARI
72
F(ilius) (Vila Real, Callaecia Bracarensis). Interestingly, a
number of Continental Celtic names show a structure of
prefix + -tro-, which indicates that this -tro- must be the
verbal form governing the adverbial and the compound is
synthetic, unless -tro- is a variant of the comparative suffix
-tero- and the prefix is an adverb (cf. IE *h1en-tero-, *h1 en-
tMh2o-, etc.), which remains an alternative but less
convincing possibility in virtually every case:
L(ucius) ANTRONIVS P(ubli) F(ilius) (Narbonne); the
PN VATRICVS (southern Burgos, Arevaci), probably related
to VATRI (Venetia et Histria), to the dat. sing. VATRO
(Mallorca) and perhaps to the gentilic VATRONIVS (Italy); 73
VERTROS (Lugdunensis, Quimper) and NOREIA VERTRON(i)
BVTTON(i) / S(uo) F(ilio) ET / TETRIAE ADNOMATI / F(iliae)
S(uae) (Noricum, Friesach, CIL III, 3819); the gentilic
Autronius must also be taken into account.
I am inclined to see these forms, or at least a small
nucleus of them, as inherited prefixed synthetic
compounds of the structure *÷-t®h2ó-, *(h1)upo-t®h2ó-,
*(h1)uper-t®h2ó-,*h2eu-t®h2ó-, perhaps even *trei-t®h2ó- in the
hápax TETRIAE. 74 If they are old enough, they will have

in Celtic). They correctly adduce a Hispanic PN VLATICI (Cáceres), but


associate our divinity with VLEDICO (Rioja) which in their view is related
with MW. gwledig from *uledicos (in fact it goes back to *ulatiko-). They
translate 'el señor Sadu' or 'el señor de la paz'.
72
Delamarre’s only example of this formation (2007: 73) is VICTORIANVS
CONTARIVS in Pannonia; but this is a Latin word for ‘pike bearer’ in a
Latin onomastic context.
73
Contrary to expectations, the sequence -tr- has not been broken by a
prop-vowel -u-, which points to this sequence being tautosyllabic and the
whole word being consequently a compound. Cf. K.1.3 muturiskum (not
†mutiriskum /mutrisku:m/) < *mut-ro- ‘dark’ (cf. OIr. mothar ‘swamp’),
K.1.3 Statulu < *sta-tlo- ‘staying place’ (STATLIAE in Narbonne, but
STATVLICI in Castelo Branco, Portugal), CABVRVS < *kaf-ro-, K.1.3. Abulu <
*af-lo-, CATVRVS < *kat-ro-.
74
This name might be related to the unexplained Latin adjective tetricus
‘stern, frowning’ (originally designating ‘one who has a very penetrating

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 39

regularly lost their laryngeal without a trace, as in the case


of the -gno- compounds going back to *g÷h1o- (Gaul.
Cintugnus, Lat. benignus, Gk. neognÒw). Accordingly, -taros
compounds are secondarily created in analogy to the
simplex, and consequently more transparent.

8. Lusitanian names containing IE /p/ again. The wide


fields from which the remains of acrostatic nouns in the west
of the Roman Empire are gleaned
Two new votive inscriptions have recently come to
light in Alcains (Castelo Branco, Portugal), enriching the
vast list of unparalleled divinities and, in the first case, the
flimsy evidence for the existence of gentilitates in western
Hispania:75

1) ASIDIAE / GENTILITAS / POLTVRICIO/RVM · EX VOTO POLTV/RI


CAENO/NIS / A(nimo) · L(ibens) · S(olvit) ‘To Asidia, the
gentilitas of the Polturicii, fulfilling the promise made by
Polturus, son of Caeno’
2) ASIDIAE / POLTV/ CEAE / L(ucius) ATTIVS / VEGETVS / V(otum)
· L(ibens) · S(olvit) ‘To Asidia Poltucea, L. Attius Vegetus’

As the editors correctly remark, ASIDIAE can be


directly related to the epithet of the divine name in an
inscription reading NIMMEDO / [A]SEDDIAGO / G(aius).
SVLPICIVS / AFRICANVS / V (otum) S(olvit) L(ibens) M(erito)
(Mieres, Asturias), to which may be added
DERCINOASSEDENSIBVS (Montejo de Tiermes, Soria),
Gaulish ASSEDI, ASSEDOMARI, etc. (Delamarre 2003: 33). It
goes back to *ad-sed-iá, meaning ‘resident’ (cf. Lat.
assiduus).
There is yet no explanation for POLTVRI and
POLTVRICIORVM . The later form has been enlarged with a
velar suffix of the form *-ik-io-, which is not very usual in
Celtiberian family names derived from personal names, but
not unheard of in roughly the same region as a formant
deriving divine epithets from place names.76 The common

look’?) if it contained /e:/.


75
Edition by Assunção et alii (2009).
76
TANGINICIAECO (Castelo Branco), TRITIAECIO (Cáceres), MANDICEO
(Sintra), ARANTONICEO (Lisboa), ARPANICEO (Badajoz), ABERCICEA
(Cáceres).

Volume 43, Number 1 & 2, Spring/Summer 2015


40 Blanca María Prósper

base of both names is probably a very primitive adjective


*poltu-ro-. The formation of POLTVCEAE points to an
identical stem *poltu- with addition of a velar suffix; 77 bear
in mind that when a suffix -iko- gets in contact with an -u-
stem, final -u- usually predominates and even absorbs the
first vowel of the suffix (or alternatively selects a simpler
suffix). 78
A possibly identical type is reflected in the PN
MEDVCEA (Viana do Castelo) if it is not a misspelling for
the common Medugena. An inscription recently brought to
light by Olivares et alii (2009: 194-195) reads CASI/NA
DE/[O] NITV /[ C] EO V/OTVM / S(olvit), etc. (Cerezo,
Cáceres, Lusitania Emeritensis). If the divine name is well
read, NITVCEO is the derivative of Celtic *niH-tu- ‘fight’, in
MIr. níth (masc., perhaps originally an u-stem) ‘conflict;
anger, resentment’, going back to *nítV-, which is a
cognate of Skt. nítíh ‘leading’, Goth. neip ‘envy’, OE. ní
‘combat, hate, enmity’ all of them from *neiH- ‘lead, guide’
(cf. Rix et alii 2001: 450; Zair 2012: 116).
Widmer (2004: 126) has reexamined the relationship
of a number of nouns in -u- and adjectives derived from
them. In Widmer’s view, apparent Schwebeablaut must be
explained sometimes as a product of the opposite
phenomenon: an acrostatic form *polth2u- has come into
being by internal derivation from a proterokinetic
adjective and not the other way round as usually accepted.
Since we are dealing with a root *pleth2-, Widmer
reconstructs a converse process *pleth2u-/*pÒth2eu- ‘eben,
flach, breit’  *pe/olth2u- ‘Flache, Ebenheit’. Of the two IE

77
In POLTV/CEAE there is room for at least one more letter at the end of
the line after <V>, so that one wonders why the text stopped there, and in
view of the photograph we even might toy with the possibility of a letter
< R> having been erased. In that case, the divine name might be
ultimately identical to that of the gentilitas, but this is mere speculation.
78
This kind of formation may be the continuant of an ancient non-Latin
scheme, often found in Celtic PNs: Cf. the divine epithet NANTVGAICIS in
Callaecia, from *nantu-ko-, the FN CONTVCIANCO (Segobriga), the PN
MATVCVS in Gaul, from *matu--, TRETVCIO (dat. sing., hápax, Noricum),
from *treiH-tu-, related to MIr. tréith (-i-stem adj.) ‘weak, cowardly’ <
*treiti-, tríath ‘weak’ < *treito- are derived from *treiH-tV- (Irslinger 2002:
214-215); VETVCIVS (hápax, Noricum), if from *ueiH-tu-; BITVCVS (Gaul)
from *guîtu- ‘life’.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 41

acrostatic inflections, resp. -e/o- and -é/e-, Widmer (2004:


127-128) rightly favors the first on the grounds that this is
the usual type in -u-stems (*h2 e/oiu- ‘lifetime’, *kre/otu-
‘strength’, *pe/olh1u- ‘amount’).
The adjective in question is attested in Skt. p®thú-, Gk.
platÊw, and slightly remodelled in Lith. platùs. A noun
*pe/olth2u- gave PGerm. *felu-, the only extant example of
Widmer’s acrostatic inflection for this root, which is
directly attested in some inflectional cases of OE. feld. In
other Germanic languages, this form has gone over to the -
a-declension, which in Widmer’s opinion is an innovation
due to -u-stems being recessive. Summarizing, the Hispanic
forms may confirm the existence of a paradigm *pe/olth2u-
‘field’ further enlarged to give rise to a new adjective
*pe/olth2u-ro- ‘related or belonging to the field’79 and
contribute still another instance to the list of non-Celtic
words from Lusitania containing IE /p/.80 Any attempt to
derive POLTVRI or POLTVCEAE from *kuel(h1)-tu- (cf. Lat.
cultus) carries the burden of the proof: an /o/ grade is
unexpected here, and contextual rounding of original /e/
has no parallels in Celtic unless two labiovelars are
involved: limiting our scope to Hispano-Celtic, we find
u u
PETRAIOCI (PN, *k etrV-), PENTIVS, PINTAMVS (PN, *k enxtó-)
Pelendones (EN, *k elVnt-), PELLVS, PELLICO (PN, *kuello-),
u

etc. Note that a sequence /poltu/- could in principle be


the Lusitanian (and then non-Celtic) result of *kuÒ(h1)-tu-,
since in this dialect laryngeals seem to have been lost in
the string #C[+lab]ÒH.C- as I have posited for Italic (e. g. in
Lat. cultus; see Prósper forthc.-a), but from the formational

79
According to Nussbaum (1998: 52), /o/ grade in suh adjectives as Gk.
oÔrow < *h 1osu-ro- ‘good, favorable (wind)’ is symptomatic of acrostatic
inflection in the nouns they are derived from. Cf. Widmer (2004: 194).
Willi (2006: 194-195) has traced back Gk. nÒsow 'illness' to a privative
compound *÷-h 1 osu-o- 'lacking wellness'.
80
Cf. PORCOM, PORGOM ‘pig’; POEMANAE, PEMANEIECO ‘nourisher,
protector’ (< *poh 2 i-mVn-); PANDITI (cf. Lat. Panda, Oscan Pandina). See
Prósper and Villar (2009). Vine (2006: 234) interprets ópilió ‘herdsman’
as a possible ‘rustic’ Roman reflex of *aupolió << *aui-polo- < *oui-polh2o-
‘sheep-driver’ (or *oui-ph 2-lo- ‘sheep-protector’), with unrounding *ou- >
*au- confined to pretonic syllables, which suits my views on OIPAINGIAE
‘sheep protector’ perfectly well; I overlooked this example in Prósper
2012. Now as ever, forms containing /p/ are not Celtic.

Volume 43, Number 1 & 2, Spring/Summer 2015


42 Blanca María Prósper

point of view this option is less attractive.


Interestingly, a nearly synonymous and similarly built
Celtic counterpart of *polturo- may be attested in the FN
ETVRICO (Madrid, Carpetani, CIL II, 6310). This name may
go back to a very primitive adjective *peitu-ro-.81 According
to Widmer (17-78) *peitu-, which literally means ‘corn
producing ground’ (as in OIr. íath ‘field’), is the product of
exocentric internal derivation from *pitu- ‘corn, food,
nurture’ in Indo-Iranian pitú- and OIr. ith, OW. it. In sum,
the derivational chain is depicted as proterokinetic *peitu-
/*piteu- ‘Nahrung’  acrostatic *po/eitu- ‘Nahrung gebend,
tragend’. A glance at the whole evidence reveals there has
been paradigm levelling in favor of the /e/ grade variant
in Celtic.

9. LVMBIS (Lusitani Scallabitani). Back to the cave of the


nymphs
Most recently, a new votive inscription has come to
light in Longroiva (Meda, Guarda, Portugal).82 It reads:
POTITVS / REBVRRI / F(ilius) · LVMBIS / VOTVM /
SOLVIT
The divine name LVMBIS looks like the dative plural of
a thematic or feminine á-formation. It is unknown not
only in Hispania but in Latin epigraphy as a whole.
Interestingly, it is immediately reminiscent of a noun for
water divinities best attested in the western branch of
Indo-European: Latin lumpa ‘water spirit’, Oscan diumpais
(dative plural) ‘water spirits’ in the Oscan Tabula of
Agnone (see especially Poccetti 1997). These forms show a
well known d/l variation, and the Oscan form has
undergone accretion of a glide between a dental stop and -
u-. In Latin, a lumpa (since Pacuvius) designates a water
fairy, and the obviously related, hellenized lympha means

81 ?
AMMISA BENILTI AETVRIQ(um) F(ilia) (Illescas, Toledo) compromises
this explanation in so far as complete merge of the outcomes of CCelt.
/e:/, /ei/ and /ai/ cannot be substantiated.
82
Cf. Sá Coixão and Encarnação (2014). I would like to thank the latter
scholar for showing me the text of the inscription prior to its edition.
According to the editors, we are dealing with the dative plural of Latin
lumbus 'back, loin' conveying sexual nuances, which is unrealistic to say the
least.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 43

‘clear water’, but when used in the plural it designates


water nymphs, like the Oscan word: Cf. the epigraphic
evidence in LVMPHIEIS (CIL X, 6797, Ischia, Latium et
Campania) or in LYMPHIS SAC(rum) (CIL III, 6373, Salona,
Dalmatia), NYMPHIS LYMPHISQ(ue) AVGVSTIS (CIL V, 3106,
Vicetia, Venetia et Histria), LYMPHIS V S L M (CIL X 6791,
6796, Latium et Campania).83 There are three current
theories regarding these forms:
The traditional idea (which goes back to Wackernagel
1908: 218ff.) considers these forms as loanwords from the
predecessor of classical Greek nÊmfh; the borrowing took
place in Magna Graecia. This was the one and only original
form, and nasal dissimilation took place after it was
borrowed into Latin and/or the Italic languages (the
details of the transmission remain obscure).
A second account contends that lumpa and diumpais
are inner-Italic formations, at the cost of accepting that
their original Anlaut was d-, which thereupon evolved into
l- in Latin, as in *dakrumá > lacruma. This confronts some
difficulties given the lack of a plausible etymological
ascription. For instance, Richter derives these divine
names from a root *dup- with a nasal infix giving Italic
*dumpá and later dialectally *(d)iumpá 'spring, fountain',
which would form the base of Sardinian iumpare 'to spring,
jump' (1907: 440).
Hyllested (2004) again resorts to borrowing from
Greek, but has a different vision of the changes involved:
Gk. nÊmfh has a meaning ‘water goddess, nymph’ which
must have a different origin from the alternative one
‘married girl, virgin’. This scholar ingeniously contends
that this word was originally *lÊmfa and was borrowed into
Italic in this form before it underwent nasal assimilation in
Greek by analogy with nÊmfh (cf. Lat. nupta) and the latter
becomes a polysemic word. Next, he brings to bear another
word for a water nymph, allegedly attested in a Sanskrit
divine name: Rámbhá, a nymph belonging to the group of
the Apsarásas. Finally he relates Lithuanian Lãum ‘water
fairy’ from *loum-. He proposes to reconstruct *lumbh eh2 or

83
The connection and its precise nature was obvious for ancient scholars:
Cf. Paul. ex Fest. 107, 17: ‘Lymphae dictae sunt a nymphis’.

Volume 43, Number 1 & 2, Spring/Summer 2015


44 Blanca María Prósper

*lombh eh2. In the latter case, the Greek form must have
undergone Cowgill’s rule, by which /o/ yields /u/ in this
context, before it was borrowed into Italic. As pointed out
by Hyllested, *lombh eh2 could go back to the root *lembh -
(Rix et alii 411-412).
In point of fact, the Indian name Rámbhá can be in
principle traced back exclusively to *lombh eh2. This kind of
divinities is said to be dangerous, enchanting and seducing
men into rapture and insanity. And this name, as observed
by Hyllested, is inmediately suggestive of a connection
with *lembh -, best attested in Skt. rábhate ‘to seize’, which
has nasalized forms, such as rámbhati, rambhayati, held
sometimes to be secondary. Specifically, it might be an
agent noun, which exists as such in rambha- ‘support’ and
interestingly in the masculine PN Rambha-, king of the
Asuras. But this root is only attested in Indo-Iranian (and
possibly in the Greek adjective émfilafÆw ‘encompassing’),
and its being deeply entrenched in this system is
suggestive of the late, inner Indo-Iranian origin of rambha-
. Its relationship with the Italic forms is consequently not
proven. 84 On the other hand, that the original name of
water-giving deities should only have to do with their
wicked nature does not seem altogether likely. A preform
*lombh eh2, which is only compatible with the borrowing
hypothesis, at least as far as Oscan is concerned, has
another difficulty subtly overlooked in this version of the
aforementioned hypothesis: The word nÊmfh has the
specialized meaning ‘water nymph’ already in Homer and
Hesiod.85 Accordingly, a hypothetical assimilatory process
*lÊmfa > nÊmfa, if it has taken place only once, is arguably
predialectal. The foundation of the first colonies in Magna
84
He mentions a reconstructed Finno-Ugric *lumpa ‘nenuphar’ and a
number of river names as Indo-Uralic cognates. Since I am no expert in
these matters I cannot give an informed opinion, but I suppose the Finno-
Ugric form might just as well be a loanword from PIE or an IE branch.
85
See Ballentine (1904: 77ff.) for the first mentions of the nymphs in
connection with water, e. g. as spring deities in Homer (Od. 17, 205-211),
as a name of the Hyades in Hesiod, Fr. 181, in the early epigraphic
evidence, etc. Varro (De re rust. 1, 1, 6) provides the first evidence of
humans praying to the nymphs for water, and specifically for rain. Iuturna,
who was the mother of Fons, is known to have been a nymph related to
water since at least the third century BC.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 45

Graecia does not reach further back than the eighth


century, which makes it necessary to presuppose that
Greek independently preserved the original form and
passed it on both to Oscan and Primitive Latin or to only
one of these dialects, which then transmitted it to the
other in the form *lumpá. In consequence, the
relationship of the Italic divinity and its alleged, nowhere
attested Greek forerunner, must remain hypothetical. In
fact, it is now gainsaid by the Lusitanian form, which has
come to entangle the problem. These are the naked facts:
- The Lusitanian form is compatible with *lumb(h )-,
*lomb(h )-, *lump-, *lomp-. Preforms containing /o/ are not
unthinkable but slightly less likely, since raising of this
vowel before a labial is attested only to the south and
southwest of this area (PVMPI < *kuomku(e)-io- in Portalegre,
u u
PVPPID < *k od-k id in Cáceres: see Prósper and Villar 2009).
- This form is completely unknown in Celtic, not so in
Italic, a branch with which Lusitanian has been shown to
share a large number of coincidences. This of course has
never been taken to mean that populations left Italy to
end up settling down in western Hispania, but that in
prehistoric times, before Italic populations actually
traversed the Alpes and made it through the Italian
Peninsula, a branch thereof turn westwards and crossed the
Pyrénées, only to be virtually absorbed by upcoming Celtic
waves except in their westernmost realms (see Prósper
2002, passim). Accordingly, there is a distinct possibility
that LVMBIS, diumpais and Lat. lumpa are inherited (Para-
)Italic forms, since a loanword from Greek is excluded in
the case of the Lusitanian form.
- The Lusitanian form virtually eliminates any
alternative etymology of Oscan diumpais, Latin lumpa that
is based on the idea that the original Anlaut is d- (and
consequently not related to the Greek form in any way), as
has been contended by some scholars for more than a
century. A divine name *Dumpá is not an option any more.
- By the time the Greeks had come to be in touch
with the Italic populations, Latin and Sabellic or Oscan
must have been separate languages (supposing they ever
had a common ancestor, which is doubtful). Stepping
beyond the possibility of borrowing, that a Greek string
/lumpha:/ (let alone /numpha:/) should have modified

Volume 43, Number 1 & 2, Spring/Summer 2015


46 Blanca María Prósper

preexisting forms of a similar etymology is unwarranted.


Given that the oldest Greek texts never bear testimony to
a form *lÊmfh, this reconstructed word may have existed,
if at all, in times too early to have had any direct bearing
on the problem posed by the Italic divine names.
- The fact that these divinities are adored as a group is
unlikely to be accidental, nor can it be put down to Greek
influence of any sort. The Lusitanian form is probably
related not only because of its form or its appearing in the
slot of divine names, but also because it is a plural form like
diumpais, generically referring to the spirits haunting the
waters. Accordingly, we may depart from a single Italic form
not borrowed from Greek.
It could be argued that the original form is *lumbh á.
Since it cannot be explained as a loanword from Greek
anymore, this is not really likely, because we would expect
a regular result †lumbá. Provided that the Italic forms have
been influenced by a synchronic Greek *numph á ,
however, their consonantism may have been affected, but
this is mere speculation as implied above. The phonotactic
peculiarities of Indo-European make some difficulties for a
structure *lumbh eh2 instead of regular *luMbh eh2, but these
are not unsurmountable in view of the unusual sequence
of three sonorants, in which the syllabic nucleus may have
been shifted early. It would pose no problems if it went
back to a root *leubh - ‘to seduce, be lovely, entice’, either
with a nasal suffix, which is unwarranted, or with an early
metathesis from *lubh -neh2.86
Finally, a tantalizing possibility must still be taken into
account, however farfetched it might seem at first sight,
because it seamlessly accounts for the phonetic side: That
we are dealing with an aquatic word *lumpá, which for the
time being is only known in Italic and Lusitanian, where
voicing of a stop in this context is not unexpected, and
perhaps also in Baltic, if Baltic river names like Lithuanian
Lum̃p are added to the list (note that Hyllested treats it as
Finno-Ugric, but Lithuanian river names are mostly of
Indo-European origin and witness to the largest number of

86
This form underlies Albanian (Tosk.) lumë, (Geg.) lum ‘happy’, lumní
‘happiness, glory’.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 47

aquatic roots and formants). If it originally meant ‘water’ it


could go back to a nasalized variant of the aquatic root
*leup-.
*Lumpá may have influenced Gk. nÊmfh < *nubh neh2
through the colonies of Magna Graecia (or through
commercial contacts reaching back as far as the Mycenaean
age) when its pronunciation was *numph á, thus giving rise
to its second meaning (unless, of course, the same *lumpá
had been inherited by Greek and assimilation took place
without external influence), but once more there is no
way of substantiating this point. Later on, Lat. Lumpa was
glossed over as Lympha in cultivated environments.
Contrary to what is implied by the two current hypotheses
involving borrowing, this does not necessitate a two-stage
process like that of Lat. ampulla vs. later amphora.
Such forms as Lat. limpor ‘a clear liquid’, limpidus
‘clear’ seem formally unrelated and contain an original
voiceless labiovelar, but it has been noted that lumpa
alternates with limpa 'water, water spirit', itself probably the
product of contamination with limpor. Accordingly, this
group of words may have played a role in the devoicing of
the labial (if we depart from *lumbh á) and especially in the
attraction of lumpa to the Greek word, which by that time
already contained an [û] which later evolved into [i].87
Interestingly, both divinities are not confused in Latin,
where they occur side by side in the same text in CIL V,
3106 (Vicetia) and in Augustine, De civitate Dei 4, 34.
Especially illustrative is the word play in an inscription
reading [....] VT LIMPHIS EIVS LIMPIDIS AC DVLCIBVS /
LEVENTVR MAGNAE MOLESTIAE ET CVRAE / CALP{ H}VRNIVS
L(uci) F(ilius) (Aletrium, Latium et Campania, cf. Garofoli
2013: 131). If these forms were the product of a repeated
process of borrowing, we would probably find lumpa and
nympha only, since no Greek model for lympha was available
when the transcription <y>, <ph> started to be used around
the 2 nd century BC. The intermediate stages lumphieis,
lymphis are only diagnostic of a semantic connection

87
Hyllested is likely to be right when he takes limpa to be the product of
crossing and not an original form as opposed to secondary lumpa, as held
by Solta (1967: 95). This scholar traced limpa, limpor back to a present
*limp-é- comparable to Lat. liqueo and linquo.

Volume 43, Number 1 & 2, Spring/Summer 2015


48 Blanca María Prósper

between Lumpa and nympha that may have nothing to do


with their form.

References

Abascal Palazón, José Manuel


2011 Airones y aguas sagradas. In: A. Costa, Ll. Palahí and D. Vivó
(eds.), Aquae Sacrae. Agua y sacralidad en la Antigüedad, 249-
256. Gerona.

Abásolo Álvarez, José Antonio and Albertos Firmat, María Lourdes


1976 Acerca de unas inscripciones de Poza de la Sal. Boletín de la
Sociedad Española de Amigos de la Arqueología 42: 393-407.

Abásolo Álvarez, José Antonio and Rodríguez, Adelaida


2005 Caleca Nevara. In: L. S. Iglesias Rouco, R. J. Payo Hernanz and
M. P. Alonso Abad (eds.), Estudios de Historia y Arte. Homenaje
al Profesor D. Alberto C. Ibáñez Pérez, 107-112. Burgos.

Adams, Douglas Q.
1999 A Dictionary of Tocharian B. Amsterdam: Rodopi.

Adams, James Noel


2007 The Regional Diversification of Latin, 200 BC - AD 600.
Cambridge: Cambridge University Press.

Albertos Firmat, M. Lourdes


1952 Nuevas divinidades de la antigua Hispania. Zephyrus 3: 49-63.

Álvarez González, Yolanda, López González, Luis, López Marcos, Miguel


and López Barja, Pedro
2004 Dos inscripciones inéditas del castro de San Cibrán de Las
(San Amaro-Punxín, Ourense). Palaeohispanica 4: 235-44.

Assunção, António, Encarnação, José D’. and Guerra, Amílcar


2009 Duas aras votivas em Alcains. Revista Portuguesa de Arqueologia
12: 177-89.

Ballentine, Floyd G.
1904 Some phases of the cult of the Nymphs. Harvard Studies in
Classical Philology 15: 77-119.

Beekes, Robert S. P.
2003 Historical phonology of Classical Armenian. In: F. Kortlandt
(ed.) Armeniaca, comparative notes, 133-211. Ann Arbor.
2010 Etymological Dictionary of Greek I-II. Leiden: Brill.

Béjar Trancón, María Belén


1995 Cuatro nuevas estelas de la provincia de Burgos. Boletín de la

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 49

Sociedad de Amigos de la Arqueología 61: 197-207.

Beltrán Lloris, Francisco and Díaz Ariño, Borja


2007 Altares con teónimos hispano-célticos de la Meseta norte
(Museos de Palencia, Burgos y Valladolid). In: M. Hainsmann
(ed.), Auf den Spuren keltischer Götterverehrung. Akten des 5.
F.E.R.C.A.N.-Workshop, 29-56. Vienna.

Bernardo Stempel, Patrizia De


1999 Nominale Wortbildung des älteren Irischen. Tübingen: Niemeyer.

Birley, Anthony R.
1980 The People of Roman Britain. California: University of
California Press.

Bozzone, Chiara and Felisari, Clara


2012 Initial “Jod” in Greek and the etymology of Gk. ·ppow
‘horse’. Talk held at the 24th Annual UCLA Indo-European
Conference, Los Angeles: 1-6. On line on the internet:
www.academia.edu.

Carcedo de Andrés, Bruno Pedro


2008 Religiones prerromanas en la Hispania antigua. Los
turmogos. Norba 21: 159-79.

Castillo, Carmen, Gómez-Pantoja, Joaquín and Mauleón, María Dolores


1981 Inscripciones romanas del Museo de Navarra. Pamplona:
Diputación Foral de Navarra.

Delamarre, Xavier
2003 Dictionnaire de la langue gauloise. Paris: Errance.

Demicheli, Dino, Radman-Livaja, Ivan and Wiewegh, Zoran


2012 List of personal names on a ceramic bowl fragment from
Siscia. In: Michel E. Fuchs, Robert Sylvestre and Christophe
Schmidt Heidenreich (eds.), Inscriptions mineures: nouveautés et
réflexions, 185-192. Bern.

Dunkel, George
2007 Chips from an aptotologist's workshop I: 1. Departiculative
thorn, 2. The locatives of *kás- ‘series, row’, 3. Emphatic
*éh 1 /óh 1 'so' and sentence-initial a- in Anatolian, 4. Radices
departiculativae. In Alan J. Nussbaum (ed.), Verba Docenti.
Studies in historical and Indo-European linguistics. presented to Jay
H. Jasanoff by students, colleagues, and friends, 53-61. Ann
Arbor.

Elorza Guinea, Juan Carlos and Abásolo Álvarez, José Antonio


1974 Nuevos teónimos de época romana en el País Vasco-Navarro.
Estudios de Arqueología Alavesa 6: 247-258.

Volume 43, Number 1 & 2, Spring/Summer 2015


50 Blanca María Prósper

Emmerick, Ronald
1992 Indo-Iranian. In: Jadranka Gvozdanovic (ed.), Indo-European
numerals, 163-198. Berlin-New York.

Encarnação, José D’
2009 Duos monumentos epigraficos da civitas Igaeditana. Praça
Velha. Revista cultural da cidade da Guarda 26: 162-171.

Fleuriot, Léon
1974 Sur quelques textes gaulois. Études Celtiques 14: 57-66.

Garofoli, Pietro
2013 La dedicata alle "Nymphae hospites" di Guarcino. Epigraphica
75: 127-140.

Garcia, José Manoel


1991 Religiões antigas de Portugal. Lisbon: Imprensa Nacional-Casa
da Moeda.

García Alonso, Juan Luis


2010 De etimología y onomástica. Deo Aironi y Pozo Airón.
Palaeohispanica 10: 551-566.

Gómez-Pantoja, Joaquín, Madruga, José Vidal and González Cordero,


Ángel
2013 Cuatro altares de la Vera, Cáceres. Archivo Español de
Arqueología 86: 279-292.

Hayes, Bruce
1989 Compensatory lengthening in moraic phonology. Linguistic
Inquiry 20: 253-306.

Hispania Epigraphica. Archivo Epigráfico de Hispania. Madrid,


Universidad Complutense de Madrid. Online edition:
http://eda-bea.es/.

Hyllested, Adam
2004 L'esprit des eaux: grec nymphé, sanskrit Rámbhá, lituanien
lãum et quelques autres formes semblant apparentées. In: A.
Hyllested, A. R. Jørgensen, J. H. Larsson and Th. Olander
(eds): Per Aspera ad Asteriscos. Studia indogermanica in honorem
Jens Elmegard Rasmussen sexagenarii, 219-233. Innsbruck.

Irslinger, Britta S.
2002 Abstrakta mit Dentalsuffixen im Altirischen. Heidelberg:
Winter.

Jasanoff, Jay
1988 Old Irish boí ‘was’. In: M. Jazayeri and W. Winter (eds.),

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 51

Languages and Cultures. Studies in Honor of Edgar C. Polomé,


299-308. Berlin-New York.
2003 Stative -é- revisited. Die Sprache 43: 127-70.

Jordán Cólera, Carlos


2005 ¿Sistema dual de escritura en celtibérico?. Palaeohispanica 5:
1013-1030.

Joseph, Lionel S.
1982 The treatment of *CR H- and the origin of CaRa- in Celtic.
Ériu 33: 31-57.

Kavitskaya, Daria
2002 Compensatory Lengthening. Phonetics, Phonology, Diachrony.
London-New York: Routledge.

Kroonen, Guus
2013 Etymological Dictionary of Proto-Germanic, Leiden: Brill.

Lazzarini, Maria Letizia and Poccetti, Paolo


2001 Il mondo enotrio tra VI e V secolo a.C. L’iscrizione paleoitalica da
Tortora. Naples: Loffredo.

Lubotsky, Alexander
1994 The original paradigm of the Tocharian word for king.
Tocharian and Indo-European Studies, Suppl. 4: 66-72.

Marco Simón, Francisco


1993 Nemedus Augustus. In: I. Adiego, J. Siles and J. Velaza (eds.),
Aurea Saecula. Studia Palaeohispanica et Indogermanica J.
Untermann ab Amicis Hispanicis Oblata, Barcelona: 165-177.

Martirosyan, Hrach R.
2009 Etymological Dictionary of the Armenian Inherited Lexicon. Leiden:
Brill.

Matasovic, Ranko
2009 Etymological Dictionary of Proto-Celtic. Leiden: Brill.

Mayer, Marc and Abásolo, José Antonio


1997 Inscripciones latinas. In: J. A. Abásolo (ed.), La Cueva de la
Griega de Pedraza (Segovia), 183-259. Zamora.

McCone, Kim
1991 The Indo-European origins of the Old Irish nasal presents, subjunctives
and futures. Innsbruck: Innsbrucker Beitrage zur
Sprachwissenschaft.

Volume 43, Number 1 & 2, Spring/Summer 2015


52 Blanca María Prósper

1996 Towards a relative chronology of Ancient and Medieval Celtic sound


change. Maynooth: National University of Ireland.

Nussbaum, Alan J.
1976 Umbrian pisher. Glotta 54: 241-253.
1986 Head and Horn in Indo-European. Berlin-New York: Walter de
Gruyter.
1996 Latin acétum, aurítus, acútus, avítus: Four of a kind?, Talk held
at the fifteenth ECIEC, Yale University (New Haven).
1997 A note on Hesychian t°ru and t°ruaw. In: D. Q. Adams (ed.),
Festschrift E. P. Hamp II, Journal of Indo-European Studies
Monograph 25, 110-119. Washington.
1998 Two Studies in Greek and Homeric Linguistics. Göttingen:
Vandenhoeck & Ruprecht.
2010 PIE -Cmn- and Greek trnÆw ‘clear’. In: R. Kim, N. Oettinger,
E. Rieken and M. Weiss (eds.), Ex Anatolia Lux. Anatolian and
Indo-European studies in honour of H. C. Melchert, 269-277. Ann
Arbor.

Olivares Pedreño, Juan Carlos and Río-Miranda Alcón, Jaime


2009 Tres aras votivas procedentes de Cerezo (Cáceres) y una
nueva gentilitas en Lusitania. Lucentum 28: 193-200.

Olive, Joseph P., Greenwood, Alice and Coleman, John


1993 The Acoustics of American English Speech: A Kinetic Approach.
New York: Springer.

Orel, Vladimir
2003 A Handbook of Germanic Etymology. Leiden: Brill.

Pike, Moss
2011 Latin -tás and related forms. Unpublished diss., University of
California, Los Angeles.

Poccetti, Paolo
1997 Culti delle acque e stadi della vita muliebre: dottrine
misteriche e fondo religioso italico nella tavola osca di
Agnone. In: L. del Tutto Palma (ed.), La Tavola di Agnone nel
contesto italico, 219-242. Florence.

Pokorny, Julius
1959 Indogermanisches Etymologisches Wörterbuch. Bern: Francke.

Prósper, Blanca María


2002 Lenguas y religiones prerromanas del Occidente de la Península
Ibérica. Salamanca: Eds. Universidad de Salamanca.
2005 Estudios sobre la fonética y la morfología del celtibérico. In:
F. Villar and B. M. Prósper, Vascos, celtas e indoeuropeos: Genes y
lenguas, 153-364. Salamanca: Eds. Universidad de Salamanca.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 53

2007 Estudio lingüístico del plomo celtibérico de Iniesta. Salamanca: Eds.


Universidad de Salamanca.
2008a El bronce celtibérico de Botorrita I. Rome - Pisa: Fabrizio Serra
Editore.
2008b En los márgenes de la lingüística celta: Los etnónimos del
Noroeste de la Península Ibérica y una ley fonética del
hispano-celta occidental. Palaeohispanica 8: 35-54.
2012 Indo-European divinities that protected livestock and the
persistence of cross-linguistic semantic paradigms: Dea
Oipaingia. Journal of Indo-European Studies 40: 46-58.
2013a Sifting the evidence: New interpretations on Celtic and Non-
Celtic personal names of western Hispania in the light of
phonetics, composition and suffixation. In: J. L. García Alonso
(ed.), Continental Celtic word formation. The onomastic evidence,
181-200. Salamanca.
2013b The enigma of benkota, the Celtic decades and the coinage of
the Ebro valley. Die Sprache 50: 1-30.
2014a Time for Celtiberian dialectology: Celtiberian syllabic
structure and the interpretation of the bronze tablet from
Torrijo del Campo, Teruel (Spain). Keltische Forschungen 6:
115-155.
2014b Some notes on the structure and meaning of the bronze ‘Res’.
Keltische Forschungen 6: 157-164.
forthc.-a Some thoughts on Proto-Italic laryngeals in the context CÒHC-
and new Italic and Celtic etymological connections. Studia
Indogermanica Lodziensia.
forthc.-b The Indo-European ordinal numerals ‘fourth’ and ‘fifth’ and
the reconstruction of the Celtic and Italic numeral systems.
Sprache 51.

Prósper, Blanca María and Villar, Francisco


2009 Nueva inscripción lusitana procedente de Portalegre. Emerita
77: 1-32.

Quintero Atauri, Pelayo


1913 Uclés. Excavaciones efectuadas en distintas épocas y noticia de
algunas antigüedades. Cádiz: Álvarez.

Radman-Livaja, Ivan
2010 Les plombs inscrits de Siscia. Diss., Paris. On line on the internet:
www.academia.edu.

Redentor, Armando
2006 Manifestações religiosas e onomástica na Civitas Zoelarum.
Conimbriga 45: 253-73.

Richter, Elise
1907 Jumpare. Zeitschrift für Romanische Philologie 31: 432-452.

Volume 43, Number 1 & 2, Spring/Summer 2015


54 Blanca María Prósper

Rix, Helmut
1992 2 Historische Grammatik des Griechischen. Darmstadt:
Wissenschaftliche Buchgesellschaft.

Rix, Helmut et alii


2001 2 Lexicon der indogermanischen Verben. Die Wurzeln und ihre
Primärstammbildungen. Wiesbaden: Reichert.

Sá Coixão, Antonio N. and Encarnação, J. D'


2014 Ara votiva de Longroiva (Meda). Ficheiro Epigráfico 114: 489.

Scheungraber, Corinna
2011 Keltische Personennamen in Noricum -Benennungsmotive
und Semantik. In: Das Menschenbild bei den Indogermanen.
Arbeitstagung der Indogermanischen Gesellschaft, Halle-
Wittenberg, 30-3-2011/2-4-2011.

Schrijver, Peter
1995 Studies in British Celtic Historical Phonology. Amsterdam-Atlanta:
Rodopi.

Schumacher, Stefan
2004 Die keltischen Primärverben. Ein vergleichendes, etymologisches und
morphologisches Lexikon. Innsbruck: Innsbrucker Beiträge zur
Sprachwissenschaft.
2012 Mittelkymrisch tra, tros, traws, traw, Altirisch trá und
Verwandtes. In: V. Sadovski and D. Stifter (eds.), Iranistische
und indogermanistische Beiträge in memoriam Jochem Schindler
1944-1954, 361-735. Vienna.

Sen, Ranjan
2012 Reconstructing phonological change: Duration and syllable
structure in Latin vowel reduction. Phonology 29: 465-504.

Sims-Williams, Patrick
2003 The Celtic Inscriptions of Britain: Phonology and Chronology, c.
400-1200. Oxford: Blackwell.

Solta, Georg R.
1967 Lat. limpidus und seine Verwandten. In: W. Meid (ed.), Beiträge
zur Indogermanistik und Keltologie. Festschrift für Julius Pokorny
zum 80. Geburtstag gewidmet, 93-97. Innsbruck.

Stifter, David
2006 Contributions to Celtiberian etymology II. Palaeohispanica 6:
237-245.

Uhlich, Jürgen
1989 DOV(A)- and lenited -B- in Ogam. Ériu 40: 129-133.

The Journal of Indo-European Studies


Celtic and Non-Celtic Divinities from Ancient Hispania 55

Untermann, Jürgen
1997 Monumenta Linguarum Hispanicarum IV. Wiesbaden:
Harrassowitz.
2000 Wörterbuch des Oskisch-Umbrischen. Heidelberg: Carl Winter
Verlag.

Vaan, Michiel de
2008 Etymological Dictionary of Latin and the Other Italic Languages.
Leiden: Brill.
2009 The derivational history of Greek ·ppow and ppeÊw. Journal of
Indo-European Studies 37: 198-213.

Vennemann, Theo
1988 Preference Laws for Syllable Structure and the Explanation of Sound
Change. Berlin - New York - Amsterdam: Mouton De Gruyter.

Villar, Francisco
1997 An Outline of Celtiberian grammar. Innsbruck: Innsbrucker
Beiträge zur Sprachwissenschaft.

Vine, Brent
1999 Greek =¤za ‘root’ and Schwa Secundum. In: V. Ivanov and B.
Vine (eds.), UCLA Indo-European Studies I, 5-30. Los Angeles.
2006 On Thumeysen-Havet’s Law in Latin and Italic. Historische
Sprachforschung 119: 211-249.
2008 On the etymology of Latin tranquillus ‘calm’. International
Journal of Diachronic Linguistics and Linguistic Reconstruction 5:
1-24.

Wackernagel, Josef
1908 Qua - qua. Lympha. Eruptum = ereptum. Archiv für lateinische
Lexikographie und Grammatik 15: 213-221.

Weiss, Michael
2010 Observations on the littera-rule, talk held at the Cornell
University: 19-6-2010. On line on the internet:
www.academia.edu.

Widmer, Paul
2004 Das Korn des weiten Feldes: Interne Derivation, Derivationskette
und Flexionsklassen-hierarchie. Aspekte der nominalen Wortbildung
im Urindo-germanischen. Innsbruck: Innsbrucker Beiträge zur
Sprachwissenschaft.

Willi, Andreas
2006 Unholy Diseases: Or why Agamemnon and Tuthaliya should
not have offended the gods. In: D. Kolligan and R. Sen (eds.),
Oxford University Working Papers in Linguistics, Philology &
Phonetics 11: 190206.

Volume 43, Number 1 & 2, Spring/Summer 2015


56 Blanca María Prósper

Wodtko, Dagmar S., Irslinger, Britta and Schneider, Carolin


2008 Nomina im Indogermanischen Lexikon. Heidelberg: Winter.

Zair, Nicholas
2012 The Reflexes of the Proto-Indo-European Laryngeals in Celtic.
Leiden: Brill.

The Journal of Indo-European Studies

You might also like