The Instrumental Case in The Thematic No
The Instrumental Case in The Thematic No
The Instrumental Case in The Thematic No
of Continental Celtic1
–––––––
1
This work has been financed by the Spanish Government (Project DGCYT,
FFI2008–03252: Lenguas y Pueblos en la Prehistoria de Europa y Asia Suroccidental a
la luz de la genetica y la linguistica). The notation used here is the same as established
in F. Villar (1996), where the traditional distinction <ś> vs <s>, inherited from the
traditional transcription of the corresponding Iberian graphs, is replaced by <s> vs <z>.
The Celtiberian texts will be quoted according to MLH IV as far as possible.
2
Cf. C. García Merino – M. L. Albertos (1981).
3
Cf. C. García Merino – J. Untermann (2002)
4
He further argues that the simpler as opposed to the more complex signs for <ta>
have been intentionally used to tell the voiceless from the voiced consonants, which in
his view favours a transcription taruodureska dureita esainis gortika usama antos
saikios baisais kaldaikikos. This so called “dual system”, that is not autochthonous but
goes back to a north–eastern variant of the Iberian script that secondarily introduces a
means of telling voiced from voiceless stops (or, according to some scholars, preserves
the original distinctions), might have something to recommend itself, but so far it applies
only to some western documents, especially this one and Cortonum (K.0.7), and it looks
like its use was not completely consistent or didn’t have doublets for every symbol in
every document.
On the other hand, these four people are mentioned without a proper
filiation formula. This is not unthinkable, given the lack of space in the
document, and some parallels can be brought up. We could even inter-
pret Baisais Kaltaikikos as a typical formula of Personal Name + Family
Name, admitting that Kaltaikikos concords with Baisais where we
would reasonably expect a family name in the gen.pl. †Kaltaikikum.
Antos Saikios could have a similar structure, with the evident caveats
inherent to the non–existence of a Celtiberian suffix –(a)iʢk–iʢo– playing
this or any other role except for the place name Sarnikio, Sarnikiei
(K.1.1). Still, the overall impression is not satisfactory.
And what about the word Usama that immediately follows the head-
ing and precedes the alleged personal names? Since the inscription was
actually found in the Cerro del Castro, the ancient Uxama Argaela of
the Arevacī, very close to the present town called El Burgo de Osma
(Soria), which inherits the ancient name, there is little doubt that Usama
refers to this well–known place, and is not a personal name. It regularly
continues CC. *uχsamā “highest” and is thus reminiscent of other an-
cient Celtic place names like Gaulish Οὐξισάµη and Hispanic–Celtic
Οὐάµα/VAMENSI (Badajoz, Baeturia Celtica), as well as its namesake
Uxama Barca (Osma de Valdegovía, Álava) and modern names like
Osmo (Orense), Ulzama (Navarra, in medieval documents Utzama).
–––––––
deceased person is said to be an VX(AMENSI?). See below.
11
A dissimilatory tendency of the geminate sonorants –rr–, –nn–, –ll– to become –rd–,
–nd–, –ld– can be inferred from a number of personal names: APLONDVS (Madrid,
Badajoz) for APLONVS; perhaps Turenta (K.1.3) stands for Turenna, attested in
western Hispania; its only parallel is the Lusitanian TVRENDI (Castelo Branco; cor-
rected reading by AE 1984, 478); CAESARDIA (León, said to be a Viminaciensis and
thus from the area of the Vaccaeī) beside CAESARRIA (Burgos), where the geminate is
due to secondary gemination in the segment *–r.iʢ– > –r.riʢ–, and in the Romance
languages Sp. izquierda “left”, from Basque ezkerra, ardilla “squirrel”, OSp. libeldo
from libellum, Sp. celda from Lat. cella, bula/bulda from Lat. bulla. The change is also
known to affect the Aragonese language, where there is marrano/mardano “lamb”,
barro/bardo “mud”, etc. Cf. A. Zamora Vicente (1989, p. 234). In the epigraphic
evidence the dissimilation may be due to the erroneous belief, on the part of the scribe,
that the segment –rr–, –nn– in the mouth of foreigners is a corruption of regular –rd–,
–nd–.
Jasanoff himself, like Lat. calefaciō “to heat”, OI. śulā kṛ- “to roast with
a skewer”.
This archaic ending *–ōiʢs seems to have been completely ousted by
*–bhi(s) in the Celtic thematic stems, as seems to be the general ten-
dency in Indo–Iranian. Yet, P.–Y. Lambert (1994, p. 55) has tracked
down an instr.pl. in *–ōiʢs in Gallo–Greek forms like τοουτιους “avec
ses concitoyens” (Vaison), and possibly also in later texts, as in
VINDVLVS (La Graufesenque), etc. In spite of the syntactic dif-
ficulties, this is not implausible, since the Gallo–Greek texts are admit-
tedly older. Thus, –ūs, with loss of the offglide as in the dative singular
–ū, would have been progressively replaced by –obi, partly owing to the
erstwhile instrumental ending’s overlapping with the accusative plural;
and the allomorphy may have been resolved earlier in some regions than
in others. For the time being, no instrumental plural is attested in
Celtiberian. For the masculine, we would expect an ending †–uis, or †–
obi(s) if the ending had already been analogically drawn from the
consonant stems before the 2nd C. b.C.
We may well wonder if Celtic had inherited both IE *–bhi and *–bhis
(or *–bhīs from *–bhi-is, following H. C. Melchert and N. Oettinger’s
recent version of Jasanoff’s idea). In contradistinction to other IE
languages, lack of final –s in Gaulish –bi, –be does not seem to have
made this ending either adopt singular meaning or become indifferent to
number. In fact, its plural meaning must be the reason for the analogic
loss of –s in the dative plural from the oldest documents, witness Gaul.
µατρεβο, ναµαυσικαβο, etc., which cannot go back to CC., since the rest
of Continental Celtic seems to show nearly unanimously –bos: Lepontic
ariuonePos, Celtib. areikoratikubos, Noricum VIBEBOS (under the
influence of Venetic?).
Some new evidence from Western Hispanic-Celtic is most intriguing:
An inscription from Viseu (Lusitania) reads DEIBABOR IGO
DEIBOBOR VISSEIAICOBOR APINVS CHAEREAE F(ILIVS) V S L
M, which of course is nothing but the indigenous pendant of the Latin
formula deis et deabus (Viseaecis).15 This rhotacized variant of –bos,
that reminds of the destiny of final –s in Umbrian or West Germanic,
alternates with shorter –bo in cases like the dedications to LVGVBO
ARQVIENOBO vs LVCOVBVS ARQVIENIS (both in Lugo, Callaecia
Lucensis) or, more interestingly, DEIBABO NEMVCELAIGABO (Vila
–––––––
15
Edition by L. Silva Fernandes et alii (2009), who propound this on a suggestion of
mine at the Lisbon Colloquium, where they (and not only they) played around with other
implausible segmentations.
instrumental case was still alive at some point in the Ionic epic tradition,
which also accounts for the extensive preservation of -φι.18
I believe there is a chance that Celtiberian baisais is a feminine place
name in the instrumental plural case, and that its ending is the
innovative *–āiʢs of Italic and Greek. There is a number of advantages
with this interpretation: First, a welcome parallelism with Usama is
obtained. We have to do with two place names in the instrumental case,
both in turn symmetrically followed by parallel sequences in the
nominative case: antos saikios and kaltaikikos. Second, we can draw the
conclusion that the core of the document itself is usama antos saikios
baisais kaltaikikos, which explains the long, solemn and probably
stereotyped, unnecessarily long heading and dispenses with the long
sequence of otherwise unattested personal names.
The place name Baisais then must have been a plurale tantum. This is
comparatively usual (cf. Pisae, Minturnae, Phaesulae, etc. in ancient
Italy).19 Unfortunately, the ambiguities of the Iberian script prevent us
from deciding between a transcription Bais– and Pais–. Both are well
attested in southern Europe and possibly belong to a pre–Celtic
substrate, whatever the Celtiberian adaptation and pronunciation of the
initial phoneme may have been. Examples of pais–/pis– are: Pisae (to-
day Pisa, Tuscany), Pisaurum cum fluvio (Pesaro, Italy), Pisoraca (to-
day the river Pisuerga in Spain), Pisis (a river in Italy), and, with a full–
grade root *paiʢs–, the ethnonyms Paesicī (Pliny 3.28, 4.111; Ptolemy
2.6.5, a people of the Astures of northwestern Hispania) and Paesurī
(Pliny 4.113, Viseu, Lusitania), the origonym PAISICAICO (used as a
divine epithet, Viseu, Lusitania), and the place name Paisula (Ptolemy
2.4, Hispania Baetica), probably related to Phaesulae (Italy), the river
and city Paisos (Asia Minor), etc.20 The Bais– attestations are the
mostly compounded place names Baes–uri, Baes–ippo, Baes–ucci,
Besaro, all of them in the Hispania Baetica, and it seems less likely that
Baisais belongs here.21
–––––––
18
According to Prof. D. Gary Miller, in the updated version of his online course in
Ancient Greek Dialects, p. 315 (http://www.clas.ufl.edu/users/jmarks/Miller-GkDia
lects).
19
An alternative possibility is of course that it is a real plural, that is, that there was a
Celtiberian appellative †baisa, possibly referring to some feature of the landscape or to
human constructions, like the Sp. place names Las Navas, Los Montes, Las Dehesas, etc.
It might be a past participle of a verb attested in Latin as baetō “to go away” (etymology
unknown).
20
See F. Villar (2005, pp. 29–44).
21
Cf. F. Villar (2000, pp. 239–44, 283–89, 332, 409, etc.).
That of the –ā–stems, in turn, is OI. –āsu, Lith. –ose, OCS. –axƴ. Only
Greek has *–āiʢsi (alternating with *–āiʢs as a synchronic dative for
instance in Homer) beside original –āsi (reduced to the function of an
adverb, as in Ἀθήνησι) on the analogy of its masculine thematic
counterpart. Whereas the Celtiberian thematic locative singular in –eiʢ is
well attested (cf. somei, Kortonei), no locative plural has been identified
yet. Following a suggestion by F. Villar, I have contended that esokiaiz
in the ‘leaden letter’ of Iniesta (Cuenca) contains a feminine locative
plural ending in *–āiʢ–si/u that has lost its final vowel, probably some
time before Celtiberian became a written language.23
The ‘sigma’ <z> of the Iberian script is a cover symbol for three dif-
ferent sounds, probably [z], [ð] and in a vast eastern region, broadly
coincident with the ebro valley, also [ts].24 The sound [z] is the result of
the Celtiberian voicing of IE /s/, that takes place between vowels or
sonorants, but never in final position. The Celtiberian phonological
system is still far from clear, but it looks like /s/ and /z/ are distinct
phonemes in medial position (cf. the personal name Alizos from *alisos
vs the place name Letaisama, from the superlative *(φ)let(a)issamā with
expressive gemination). On the other hand, final –[z] from –sV
(wherever there has been apocope of short vowels) and final –[ð] from –
d/–t may have been neutralized at some stage.
Consequently, the IE loc.pl. *–oiʢ–si/u and the instr.pl. *–ōiʢs were no
longer distinguished in Celtiberian by the presence or absence of the last
vowel, but only by vowel length and by the opposition /z/ – /s/, since
they had probably evolved into †–oiz and †–uis before our first texts
were committed to writing; the gen.sg. and abl.sg. of ā–stems are re-
spectively –ās (Turuntas) and –āð (Kontebiaz). Now we can put forward
a hypothesis that explains the actually attested form (bear in mind that
the original length opposition is conventionally preserved throughout,
even if short and long diphthongs might have already merged):
In Proto–Celtiberian, as in Greek, an analogical proportion has taken
place: *–ūiʢs: *–āiʢs = *–oiʢzi/u: X, by which *–āzi/u is replaced by
*-āiʢzi/u or *–ăiʢzi/u. Thus, if it is ever confirmed that –aiz is a locative
plural ending, its attested form may well have been triggered by the
previous existence of the instr.pl. *–ūiʢs/–āiʢs. The ending must have
been previously analyzed as –izi/u vs –is. Alternatively, it is not very
likely that *–ōiʢs has been shortened in *–ŏiʢs in CC., even if the only
evidence for *–ūiʢs comes from the dubious Gaulish examples.
–––––––
23
Cf. B. M. Prósper (2007).
24
See the latest account in B. M. Prósper (in press–1).
–––––––
25
Possibly once in a noun stoteroi (Botorrita IV, see F. Villar et alii 2001). But this
possibility loses some ground if we admit that it might be a pronoun, too, and that
thematic nouns might have kept original –ūs from *–ōs in contrast to the rest of Celtic.
26
In K.0.7, loukaiteitubos (which in my opinion can be read loukeiteitubos), is a
fossilized double locative meaning “in the open and inside the covered areas”, from
*louʢk-eiʢ (cf. Lat. lūcus) and *teχto-bos (PIE *(s)teg- “to cover”). Cf. B. M. Prósper (in
press-2). This seems to speak in favour of the idea that the locative is undergoing
syncretism with the dative, but a dative itself is not excluded if we maintain the
traditional reading of MLH IV. Eastern and western dialects of Celtiberian may reflect
different stages of the process.
27
See F. Villar (2000) and a critical reappraisal of this hypothesis in the light of new
findings in B. M. Prósper (2007 and in press-1).
–––––––
28
Cf. B. M. Prósper (in press–2). DVRETA SALDANICA, given its insertion in a
Latin context, has traditionally been interpreted as a (hitherto unheard of) personal name
+ origonym.
29
Cf. B. M. Prósper (in press–1 and in press–2).
explicitness of the heading, neither the keyword kar nor any formula
remotely related with friendship or mutual support occurs.
In sum, I believe K.23.2 doesn’t reflect a covenant of mutual friend-
ship between two cities, Tarvodurum and Uxama, but actually
establishes the limits of their respective territories and the limits of
Tarvodurum with Baesae. I additionally believe that Baesae is to be
identified with the origin of the coins that read Kaiseza Bais, whether
this is in turn Caesada or not. If all this were true, then Tarvodurum
must have lain somewhere between Uxama and Kaiseza. Assuming
Kaiseza is identical to Caesada, the only important settlement of the
Arevacī that is more or less equidistant from these two points is Termes
(Tiermes, Soria), the ruins of which are still visited today, which could
correspond to the coinage of Tarmeskom.30 In spite of their superficial
resemblance, the identification of this place with Tarvodurum is of
course no more than a guess. If Jordán is right, we might conjecture that
the city was alternatively known by its inhabitants as a *dūrom
tarβeskom! Note that an analysis tarmes-ko- is impossible, since the
suffix –ko- as such does not exist. But tarm-esko- and tarvo-dur-eskā
show identical derivations. Bear in mind that this would probably mean
that the coins mention the actual place name, and not its derivative,
unless the place name had simply become *tarβos. 31
We have been understandably misled by the shape and size of the
bronze plaque, but this plaque, found in Uxama, may have been only the
short version of a longer and perhaps older document signed and
probably publicly shown in Tarvodurum, which may have been a more
important nucleus before Roman times, since it is Tarvodurum, not
–––––––
30
See C. Jordán (2007). Previously read bormeskom.
31
One could ascribe the different renditions of –β- to lenition processes that bring
about a graphic alternation u/b/m (see D. Stifter 2006 and B. M. Prósper in press-1). But
an alternation of a compound tarvodurum vs simple tarvos uel sim. would also be
needed. An interesting parallel can be drawn from the western H.-C. CASTELLVM
TARBV (Chaves, Vila Real), as opposed to the dedication to the LARIBVS
TARMVCENBAECIS (ibidem, from *taruʢo-okelo-bri-?, for which cf. also MARTI
TARBVCELI in Bracara). Termes would be only a Roman adaptation, like Termantia,
seemingly an invention of Appian. The reason for the creation of Termes, which is diffi-
cult to justify as a Celtic name, could be a metanalysis from the origonym Tarmestinus/
Termestinus. If this epigraphically attested form were indigenous, Tarmes/Termes could
be a Latinate back formation based on the analogy of Tibur: Tiburtinus, Ligustinus, etc.
But then Termestinus would contain in its turn a metanalyzed suffix –estino-: Cf. Lat.
domesticus, caelestis, OCS. kroměštĭnĭ “outer”, Illyrian Tergeste: Tergest-inus, but
Oneum: Onastinī. Perhaps, an ancient *tarmesc-ino- was reinterpreted by the Romans as
*tarmestino- according to their own derivational patterns and then Tarmes was
abstracted from it.
Uxama, that issues the document at all events. This document might
have included antecedents, magistrates’ names, like Botorrita, and pro-
bably references to other places or territories. Note that the fact that it is
portable would seem to make difficulties in either case: A treaty of
friendship or mutual support involving a number of cities, not a city and
a family or individual, is something nobody goes about carrying. If the
swine–shaped plaque had two sides, they may have been exhibited
respectively in Uxama and Baesae, and this might be the reason why it
was pierced after writing (although some tesserae hospitales have holes,
too).
As the reader may have guessed, the notion that this document deals
with borders is born out by the appellative noun antos, which actually
means “limit, end”. The same form and meaning is attested in OI. anta–
and the derivative PGerm. *antiiʢaz, all of them thematic forms ulti-
mately going back to IE *H2ent– “front” (cf. IEW, pp. 48–49). The only
extant Celtic parallel of antos is the accusative singular atom (with
trivial omission of the nasal; alternative reading atoš, allegedly in the
accusative plural!)32 in the Gallo–Latin bilingual from Vercelli, where
its meaning is assured by the corresponding Latin word FINIS. The text
reads:33
FINIS CAMPO QVEM DEDIT ACISIVS ARGANTOCOMATER-
ECVS COMMVNEM DEIS ET HOMINIBVS ITA VT LAPIDE[S]
IIII STATVTI SVNT
Akisios Arkatokomaterekos tošokote atom/atoš teuoxtonion eu.
The structure of K.23.2 becomes a bit more transparent now, and the
text gains in symmetry: It means that the city of Tarvodurum officially
states that “with Uxama, the border (is) the Saikios; with Baesae, the
–––––––
32
Cf. W. Meid (1989, p. 13).
33
Cf. P. Baldacci (1977–78). The reading followed by most scholars and by myself is
that of M. Lejeune (RIG II–1, E–2). More pessimistic as to the extant text is P. Solinas
(1995, pp. 381–82). Note that FINIS CAMPO instead of the expected †FINIS CAMPI
might be not a possessive expression, but one of the kind mentioned above: the bypasser
is informed that what he is approaching is “the border with the field”. The use of the
Latin dative–ablative would be due to Gaulish interference. Though eu has usually been
treated as Lat. e(x) u(oto) or a Gaulish counterpart thereof (for instance *esiʢo uʢolouʢtū
“de sua pecunia” in J. F. Eska – R. Wallace 2002), one could possibly argue for an
instrumental in –ū meaning “field” (e.g. from IE *peiʢ–uʢo– “pasture land” and
consequently roughly matching CAMPO or even *(H)eiʢ–uʢo– “passage > area?”, as in
Lith. pèr–eiva “tramp”, Goth. fraiw “lineage”, so that atoš/atom eu would more or less
mean “(A.K. has set) boundaries to access”. But one could equally argue in favour of a
syntax roughly equivalent to †FINIS QVEM A.C. CAMPO DEDIT, and eu might be a
dative singular in –ū, too: “A.K. gave a boundary / boundaries to the field”. Teuoxtonion
could be a gen. pl., as often assumed “of both gods and men”.
segienses and several Celtiberian coins read Zekia, all of which suggests
that the original place name is *Segiʢā. On the other hand, the same word
might be hidden in the first term of the compounded epithet of a divinity
BANDI SAISABRO (Portalegre, Lusitania Pacensis), if this corrupted
form goes back to a derivative of the place name *saiʢkiʢā–bri(g)–.35
References
–––––––
35
Cf. J. D’Encarnação – J. R. Correia da Silva (1994), R. M. Pedrero (2001).