You are on page 1of 16

The ‘Atlantic Fringe’ hypothesis for the Celtic homeland and the Tartessian inscriptions

By Paulus van Sluis, January 2014

Introduction

Over the last couple of decades, a series of inscriptions in a language dubbed ‘Tartessian’´
have been unearthed in south-western Iberia. These inscriptions are written in a south-western
Iberian writing system, which was imported from the Phoenician colony of Gadir. The
archaeological context of these findings allows us to date these inscriptions to the 7th and 6th
centuries BC1. The inscriptions are connected with the historical kingdom of Tartessos, which
according to Herodotos was ruled by a rich king called Argantonios between 625 and 545 BC.
This name contains the Celtic word arganto- ‘money, silver’, implying that Tartessos was
Celtic-speaking, or at least that its king had a Celtic name.2 Pre-Roman place-name evidence
of the same area gives us a mixed picture, with some 43 non-Indo-European place names in –
ippo, beside a multitude of both Indo-European and Semitic place names.3

Tartessian is still considered undeciphered by most of the relevant specialists4. Recent


efforts to interpret Tartessian have been spearheaded by John Koch, who hypothesizes that
Tartessian is a Celtic language. This assertion is controversial in that Koch has gone very far
in postulating interpretations, revisions of Celtic sound shifts, and in hypothesizing Tartession
sound shifts from a series of inscriptions in an imperfectly deciphered writing system in
which word divisions are not written. This new hypothesis is also controversial because its
inscriptions appear to date from a period several centuries earlier than other Hispano-Celtic
inscriptions. Tartessian inscriptions are even older than the oldest hitherto generally accepted
inscriptions, namely the Lepontic inscriptions attested from the 6th century BC onwards.5

1
John T. Koch, Tartessian 2: The Inscription of Mesas do Castelinho ro and the Verbal Complex.
Preliminaries to Historical Phonology, (Oxford; Oakville, CT, Oxbow Books, 2011), 1
2
John T. Koch, “Paradigm Shift? Interpreting Tartessian as Celtic” in Celtic from the West: Alternative
Perspectives from Archaeology, Genetics, Language and Literature, ed. John T. Koch and Barry Cunliffe (
Oxford; Oakville, CT, Books, 2010) 1κ9, Ranko Matasović, Etymological Dictionary of Proto-Celtic,
(Leiden; Boston, Brill, 2009) 41
3
Barry Cunliffe and John T. Koch “Introduction” in Celtic from the West: Alternative Perspectives from
Archaeology, Genetics, Language and Literature, ed. John T. Koch and Barry Cunliffe ( Oxford; Oakville,
CT, Books, 2010) 1
4
E.g. Jordán Cólera cautiously lists Tartessian as ‘Indo-European macrofamily? Celtic family?’ in Carlos
Jordán Cólera Celtibérico (Zaragoza, Ediciones del Departamento de Ciencias de la Antigüedad, 2004)
751
5
Cunliffe and Koch, “Introduction” 2
The most controversial thing about the possibility that Tartessian might be Celtic is,
however, that this view does not fit the traditional narrative of the homeland of the Celtic
languages, which is traditionally identified with the Iron-Age Hallstatt and La Tène cultures
which flourished in Central Europe from about the 8th century BCE onwards. This view is
hardly compatible with a Celtic culture all the way in southwestern Spain only a century later.
Even if one does not recognize the matrix language in Tartessian inscriptions as Celtic, the
Celtic elements in Tartessian need to be accounted for. To explain the existence of a Celtic
Tartessos from the Iron Age Hallstatt/La Tène cultures is impossible. It is simply too early. 6

Moreover, Cunliffe remarks that historical sources provide ample sources attesting to
the spread of Celts from Central Europe southwards into Italy and eastwards along the
Danube. What these sources do not provide, however, is a similar account of a spread
westwards into Gaul, the British Isles, and Spain7. Furthermore, the Central European
hypothesis dates from the days of Edward Lhuyd in the early 18th century, long before the
modern discovery of the Celtiberian language in the 20th century. Consequently, the
orthodoxy on the Celtic homeland is not designed to account for the existence of Celtic
languages in Spain. Specialists in Celtiberian have thus ever since called for a more central
position of Iberian Celts in the study of the ancient Celts.8 An often overlooked fact within the
central hypothesis, for example, is that La Tène culture never reached the Iberian Peninsula,
nor does the spread of any of these cultures explain the spread of Celtic to Ireland9. These
facts all call for a homeland hypothesis with a less peripheral role for the Hispano-Celtic
population.

Even so, relocating the Celtic homeland, whose location has not been revised for over
300 years, requires ample supporting facts and methodological rigour. For any alternative
proposal, a great deal of positive evidence and a plausible account is necessary. On the other
hand, we must not overestimate the impact of the Koch/Cunliffe hypothesis: while the
Hallstatt/La Tène cultures may be unsuitable for the spread of Celtic westwards, their spread
may still be linked to the spread of Celtic into Gaul, or to the spread of Celtic eastwards.10

6
Koch, “Paradigm Shift?” 190
7
Cunliffe and Koch, “Introduction” 13
8
Alberto J. Lorrio and G. Ruiz Zapatero, “The Celts in Iberia: An Overview”, in E-Keltoi Volume 6: The
Celts in the Iberian Peninsula (2005) 168
9
Cunliffe and Koch, “Introduction” 18
10
Cunliffe and Koch, “Introduction” 2-3
Methodology and scope

The issue this paper attempts to resolve, then, is whether or not the Celtic homeland may have
been somewhere along the Atlantic Ocean. Within the scholarly community defending this
hypothesis, there are two variant hypotheses. One is spearheaded by Barry Cunliffe, who
asserts that archaeological evidence points towards a general movement of Celtic languages
from west to east, connecting the development of Celtic with evidence of trade networks
along the Atlantic Ocean in the third millennium BC. The other school of thought is led by
Koch, arguing on linguistic grounds that Celtic developed in the Iberian Peninsula and later
spread to the northeast.11

A good methodological starting point is in deciding what is meant by ‘the Celtic


homeland’. For the purposes of this paper, the Celtic homeland refers to the area in which the
set of linguistic innovations took place that is required to reconstruct known Celtic languages
back to a Pre-Celtic state of these languages. The question that this paper attempts to answer
is in what area we can locate the linguistic innovations that gave rise to the Celtic language
group. A specific sound law often cited to define ‘Celtic’ is the loss of initial and intervocalic
*p. Alternatively, the treatment of PIE syllabic *ṛ and *ḷ as *ri and *li before a plosive may
be used as diagnostic for Celtic.12

Therefore, the question this paper aims to answer is a purely linguistic one. And to
answer a linguistic question, linguistic evidence must be used, such as language
reconstruction, place names, personal names and inscriptions. Non-linguistic evidence
stemming from fields such as archaeology must, on the other hand, be used with caution, and
cannot serve as the basis for a hypothesis on the origin of the Celtic language group. Since the
hypothesis for the Atlantic fringe as a whole seems mainly based on archaeological evidence,
Iberia will instead be the hypothetical Celtic homeland for the purpose of this paper.

The question where to place the Celtic homeland is inextricably connected to the
classification of Tartessian. If an Iberian homeland for Celtic languages is to be asserted, this
claim must be backed by demonstrating that Celtic was indeed spoken in Iberia at an early
date, and one would expect a great deal of dialectal diversity in this homeland. Conversely,
anyone seeking to demonstrate Celticity for Tartessian will need not only to provide

11
Koch, “Paradigm Shift?” 1κ5-302
12
Koch, “Paradigm Shift?” 294
interpretations of its inscriptions, but also to come up with a plausible scenario for the arrival
of Celtic in southwestern Iberia in the first half of the first millennium BC.

An overview of Indo-European in Iberia

Excluding Tartessian, the pre-Roman Indo-European languages in Iberia are Celtiberian and
Lusitanian, and several other Hispano-Celtic dialects attested through personal names in
mostly Roman sources and place names. It is still a matter of debate how these languages are
related. The chief proponent for a large number of Indo-European strata in Iberia is Antonio
Tovar. He postulated that at least one non-Celtic language with preserved Proto-Indo-
European *p (i.e. Lusitanian) entered Iberia along with several strata of Celtic. The opposite
view is defended by Untermann. He saw a single Indo-European wave of people entering
Iberia. The language of these people must have been some archaic pre-Celtic dialect which
preserved initial and intervocalic *p.13 These languages then spread out over Iberia, creating a
dialect continuum within which only Lusitanian retained aforementioned *p. This second
view is also Koch’s hypothesis, who imagines four hypothetical stages.14

1. Indo-European
2. Archaic Celtic, Para-Celtic or Hispanic: *ṛ and *ḷ become /ri/ and /li/. Includes
Lusitanian and all Celtic languages.
3. Celtic in contact with Iberian: Proto-Indo-European *p is lost in most positions.
Includes all the Celtic languages.
4. Celtic in contact with languages with /p/ beyond the Pyrenees. Includes Lepontic,
Gaulish, Brittonic and Galatian.
Francisco Villar (2004) attempts to find evidence which might support or disprove
either of these theories. He finds that, for established Celtic languages, ‘[t]he body of data at
our disposal […] is clearly compatible with the existence of a single native variety of Celtic in
the Iberian Peninsula, a variety which is relatively well known to us through the Celtiberian
epigraphy’. He attributes most of the Celtic enclaves outside the Celtiberian core area to
spread of more Celtiberians across the Iberian peninsula, stating that the majority of suggested
non-Celtiberian Celtic elements should not be considered Celtic.15 Thus, he finds no evidence
of more than one wave of Celtic migration into Iberia (apart from late incursions of Gauls into

13
Francisco Villar, “The Celtic Language of the Iberian Peninsula” in Studies in Baltic and Indo-European
Linguistics in Honor of William R. Schmalstieg Ed. Philip Baldi and Pietro U. Dini (Amsterdam, John
Benjamins, 2004) 244
14
Koch, Tartessian 2 21
15
Villar, The Celtic Language 266
the peninsula). However, he is not convinced that either Lusitanian or Tartessian bear a direct
relationship to Celtiberian, thus leaving room for Tovar’s idea that there may indeed be an
older Indo-European layer in Iberia (Alteuropäisch), bringing Indo-European into prehistoric
Iberia in multiple waves.16

The classification of Celtiberian

Celtiberian is the name given to the Celtic language that has come down to us through
inscriptions found in northeastern Spain. It is attested in a period straddling the 2nd and 1st
centuries BC. It is the language associated with various Celtic tribes in northern Iberia.
Inscriptions have been found “between the headwaters of the Duero, Tajo, Júcar and Turia
rivers and the source of the Martín River to the west, south and east, and the middle course of
the Ebro River in the north, with a frontier that runs parallel to the right bank of the Ebro,
some ten kilometres from the river, and crosses to the left bank to include an area
corresponding to a region adjacent to the border between present-day Navarre and Aragon.”17
The Celticity of this language is too obvious for comment, and its inscriptions have been
relatively confidently translated, and several grammars have been written on the language.18

Celtiberian as found in unearthed inscriptions has diverged much from traditionally


reconstructed Proto-Celtic. Celtiberian has also diverged considerably from Gaulish. While
Gaulish along with, hypothetically, Tartessian has a dative plural in –bo (e.g. ATREBO),
Celtiberian has a dative plural ending in –bos, which is shared with Lepontic.19 This example
goes to show that in the 7th-6th century BC, the Celtic language may already have developed
into separate dialects, with some of them showing –bo in the dative plural, and some of them
showing –bos.

The classification of Lusitanian

Lusitanian is the language that was spoken roughly in central Portugal and western Spain until
the 2nd century CE. It is definitely not Celtic in the traditional sense since it has preserved /p/
in PORCOM ‘pig’, for example. However, it is still undecided whether Lusitanian is closely

16
Francisco Villar, ‘Indo-Européens et Pre-Indo-Européens dans la Péninsule Ibérique’ in When Worlds
Collide : Indo-Europeans and Pre-Indo-Europeans, in T.L. Markey and J.A.C. Greppin (eds.), (Ann
Arbor, Karoma, 1990) 367ff
17
Carlos Jordán Cólera “Celtiberian” in E-Keltoi Volume 6: The Celts in the Iberian Peninsula (2007) 749
18
E.g., Jordán Cólera “Celtiberian” and Dagmar S. Wodtko., An outline of Celtiberian grammar, (Freidok,
Albert-Ludwigs-Universität Freiburg, 2003), accessed January 13, 2014, http://www.freidok.uni-
freiburg.de/volltexte/747/pdf/Celtiberian.pdf
19
Jordán Cólera “Celtiberian” ιι0-ιι2, Koch, “Paradigm Shift?” 290
related to Celtic or not, but most scholars at present reject the idea. It is only attested in a
handful of inscriptions dating from the 1st century CE, some names in Latin inscriptions, and
in some names in works of classical writers. Significantly, classical authors did not refer to
Lusitanians as either Celtiberians or celtici.20

As mentioned above, Villar does not think of Lusitanian as Celtic, thus implying that
whereas the number of Celtic migrations into Iberia was limited to one, the number of Indo-
European migrations may indeed have been higher.21 The study of Lusitanian is complicated
by the fact that it is only attested from Roman times onwards, and because by that time the
Lusitanian homeland was already colonized by Celtiberians.22 This leaves very little room to
test Koch’s hypothesis on the para-Celtic status of Lusitanian: individual Celtic words cannot
be treated as diagnostic for Celtic because they may be due to Celtiberian influence, while
specific sound laws used to diagnose Celtic may not be applicable to Lusitanian because of its
early branching off from Celtic. And this is indeed the problem of Lusitanian. Because there
is so little ground on which to establish the status of Lusitanian, it is very difficult to prove or
disprove Koch’s reconstruction of Indo-European in Iberia, making his theory slightly
unscientific.

Despite these challenges, Dagmar Wodtko (2010) has attempted to describe some
structural characteristics of Lusitanian. In a short survey she finds a series of endings looking
much like Indo-European case endings. However, context is often unable to provide sufficient
information on the specific case the endings exemplify. For endings in –a, for example, a
nominative singular of an ā-stem would be the most straightforward, but a dative, ablative
singular, or even a nominative/accusative neuter plural have all been suggested. This situation
becomes more complicated when one realizes these inscriptions date from Roman times, are
written in Latin script, and may thus even include some Latin morphological endings.23

It has been more fruitful to consider the various types of text found. It turns out that
many of these texts are dedications to indigenous divinities. These devotions typically occur
in Latin/Lusitanian bilingual texts and usually consist of the name of a deity accompanied by

20
Dagmar S. Wodtko, “Ancillary Study: The Problem of Lusitanian” in Celtic from the West: Alternative
Perspectives from Archaeology, Genetics, Language and Literature, ed. John T. Koch and Barry Cunliffe (
Oxford; Oakville, CT, Books, 2010) 335-6
21
That is: The Lusitanian corpus is analyzed, but only to see whether (proper) Celtic influence comes from a
Celtiberian source, or from another type of Celtic.
22
Villar, The Celtic Language 249
23
Wodtko, “Lusitanian” 343-344
an epithet taking the -ko derivational suffix. Sometimes only the epithet occurs. Below are
some examples of these texts, taken from Wodtko,24 with the Lusitanian in bold:

 Crougiai Toudadigoe Rufonia Seuer(a)


 Croucae Ni[.]aicui Clementinus Gei[…] a.l.u.s.
 Bandu Vordeaeco sacrum SUlpicius Paternus u(o)tum solui(t)
 …]turius Felix Vortiaeci(o) u.s.
 Nauiae Sesmacae V Anniu
 Cuhue Berralogegu ex uoto Flauius Valeri(n)u
 Coso Vdauiniago Q. V C ex uoto
 Arabo Corobelicobo Talusicobo

As can be seen, the -ko- suffix can be preceded by several vowels, but the most common
expression is –aeco- (with several variant spellings). This contrasts with Celtiberian, in which
the same suffix -ko-/-kā is used to derive adjectives from nouns, but in which the –aiko-
realization is rare. Instead, -ako-was a common suffix to form place names, while –iko-and –
oko- were generally used to form family names.25

An interesting series of phonological developments that may be shared with Celtic has
been found by Blanca Prósper by citing – hopefully – native Lusitanian divine names.26 She
cites the early developments of syllabic nasals * > aN and voiced labiovelars *gw > b in
Lusitanian in e.g. the name Bandue < *gwṃ-tu-. Syllabic resonants *Ṛ developing into Ri
cannot be proven in Lusitanian, as the only clear examples involve the Celtic –brig-.27 Prósper
also cites some developments that are usually ascribed to later Celtic. These are *kw > p and
Lusitanian dialectal development of s into h in the divine name Cossue/Cohue.28 These
developments are usually considered features of the Northern Celtic dialects of Brittonic and
Gaulish. The development of *p…kw into kw…kw has also been cited as a development shared
with Proto-Celtic.29

24
Wodtko, “Lusitanian” 345-347
25
Wodtko, “Lusitanian” 34ι-348; Francisco Villar, Estudios de celtibérico y de toponimia prerromana.
(Salamanca, Ediciones Universidad Salamanca, 1995). 126ff; Wodtko, outline of Celtiberian grammar 26
26
Blanca Ma Prósper, lenguas y religiones prerromanas del occidente de la Península Ibérica, (Salamanca,
Ediciones Universidad de Salamanca, 2002) 398, 399
27
Wodtko, “Lusitanian” 354
28
Wodtko, “Lusitanian” 353,354; Prósper, lenguas 51ff, 246ff
29
Wodtko, “Lusitanian” 360. The dating of this development is problematic, however. It is sometimes
considered an Italo-Celtic feature, but it has not spread to all Italic languages, and seems not to have fully
spread throughout Celtic languages either. See Matasović, Etymological Dictionary 11-12
Classification of Tartessian

Villar also makes separate mention of what he calls ‘Inscriptions from the Southwest’.
Referring to the Tartessian inscriptions, he, like most, is unconvinced that these inscriptions
are in a Celtic matrix language, although he does discern some elements that may point to
Celtic influence. Linguistic data on Tartessian is found in references from ancient authors,
place names and inscriptions. The Tartessian language does not have /p/, but there is a risk of
circularity here as the script has no means of disambiguating /p/ from /b/, so proof in this
matter relies on etymological conjectures.

As said before, it is known from Ancient historians that there was a Tartessian king
called Arganthonios. Even if this king did not speak a Celtic language, the fact that Tartessos
had a king with a Celtic name implies Celts had considerable prestige in southwestern Iberia.

Tartessian inscriptions are found from the early 7th century BC onwards.30 Tartessian
is written in the southwestern script. This southwestern script is primarily based on the
Phoenician script. It is a semisyllabary, in which the stop signs always carry the value of the
following vowel as well. It has furthermore added a set of vowels which are also written after
syllable signs, causing redundancy in Tartessian orthography. Untermann31 sees the addition
of vowels as evidence of Greek influence, but this does not account for the point that the
south-western script transformed Phoenician consonant signs into different vowels than Greek
did. Indeed, Rodríguez Ramos argues that the Tartessian script directly descends from the
Phoenician alephat. For example, while the Phoenician ayin provided the Greek omicron
/o/, its Tartessian descendant seems to denote /e/. A similar script is used to write Iberian,
which appears to have adopted the script from Tartessian.32 As mentioned above, Tartessian
was written in scripto continua, meaning there are no word boundaries written in Tartessian
inscriptions.

Providing a complete account of Tartessian inscriptions is beyond the scope of this


paper. Instead, we shall attempt to demonstrate how the Indo-European character of these
inscriptions can be deduced, even without knowledge of word boundaries. This is done quite
simply: whereas there are no word boundaries within an inscription, there are plenty of

30
Cunliffe and Koch, “Introduction” 2
31
Jürgen Untermann and Dagmar S. Wodtko, Monumenta Linguarum Hispanicarum 4. Die tartessischen,
keltiberischen und lusitanischen Inschriften (Wiesbaden, Ludwig Reichert, 1997)
32
Koch, “Paradigm Shift?” 203
inscriptions in which the ending is not elided. We may now seek patterns in endings and even
look for Indo-European-looking inflections or conjugations.

Here is a selection of inscriptions showing Indo-European verbal endings, taken from


Koch, with the word boundaries removed.33 The number between brackets denotes their
number as indexed in Untermann’s Monumenta Linguarum Hispanicarum 4.34

• booaranarkeenii (J.2.1)

• kiielaoe|oiśaHaabaanerobaenrkeenii (J.11.1)

• […]baebaalakiintii (J.12.4)

• bootiianakeertoorobaate’ebaarebaanaŕkeentii (J.18.1)

• akooliośnaŕkeetii (J.56.1)

As can be seen a pattern of enii/intii/entii/etii may be discerned. These must undoubtedly be a


Indo-European third person singular and plural verbal endings, with thematic inflection.35
From this point onwards, it is a safe hypothesis to assume that Tartessian was an Indo-
European language. Working from this hypothesis, it may be worthwhile to look for further
Indo-European, or perhaps even Celtic morphological devices and sound laws recurring
throughout the inscriptions.

An example of a well-known morphological device found in Tartessian is the


derivational suffix –ko/-ka. It is found in the word omuŕikaa[ (J.16.2), which, according to
Koch, stands for /womurika/ or /womorika/ ‘the under-sea place’.36 This is possibly the same
suffix as the one we find in Celtiberian to denote family names or place names.

An example of a phonological development relevant for the diagnosis of Tartessian as


a Celtic language is found in the development of the word baane, which Koch argues to be a
dative singular for the word for ‘woman’ (Old Irish ben, Proto-Indo-European *gʷenh₂) It
shows Indo-European gw having developed into Celtic b.37 The same development is found in
baaste- (S. Martinho) ‘death’ (Irish bás¸ from Proto-Indo-European *gwes- ‘extinguish’). This

33
Koch, “Paradigm Shift?” 210-257
34
Untermann and Wodtko Monumenta
35
See also Villar, The Celtic Language, 264
36
Koch, “Paradigm Shift 2ι6. Cf. Aremorica ‘place by the sea’. Koch, “Paradigm Shift 238
37
Koch, “Paradigm Shift 288
development is indeed considered Early Proto-Celtic.38 This piece of proof does of course
hinge on a correct assignment of word boundaries, but baane is a word attested several times
throughout Tartessian inscriptions.39 Evidence for the development of syllabic nasals * > aN
is found in the name Ἀ γαν ων ο /argantonios/, being based on Indo-European *h2erǵ tom
‘silver’.40 Curiously, Koch does not seem to find any examples of *ṛ and *ḷ becoming *ri and
*li before a plosive in Tartessian and mentions it no further.

Reconstructing the development Iberian Para-Celtic

Assuming Untermann is right in postulating a single Indo-European migration into Iberia, it


must be possible to reconstruct a proto-language on the basis of all the Indo-European
languages of Iberia, i.e. a Para-Celtic language including Tartessian, Lusitanian and
‘traditional’ Proto-Celtic. Since PIE *p is preserved in Lusitanian, this proto-language would
have preserved it. Instead, * > aN and *gw > b may be considered an innovation shared by
this language group.

Loss of Proto-Indo-European *p

This proto-language must then have developed into multiple dialects which preserved
PIE *p in the west and lost it in all other dialects. Notably, the non-Indo-European Iberian
language in the east of Iberia did not have phoneme /p/. Loss of Proto-Indo-European *p must
therefore, according to Koch, have been induced by the Iberian language, and therefore
constitutes proof that Celtic developed in Iberia through contact with p-less Iberian. This
assumes the plausible scenario of Iberian speakers switching to Indo-European and thus
forming Celtic.

While it is true that phonemes may be lost by such language transmission, it should be
noted that loss of /p/ is a fairly unremarkable process and does not need a substrate theory to
explain it.. Maddieson has found a pattern showing that languages with an asymmetrical
inventory of plosives often lack /p/ (and /g/). Thus he implies that it is an unremarkable
development for a language to lose /p/ from its inventory, even when such a development
causes phonological asymmetry.41 This point of criticism of the Iberian substrate-hypothesis

38
Matasović, Etymological Dictionary 7
39
i.e. J.11.1, J.19.1, J.26.1. See also Koch, “Paradigm Shift?” 274
40
This innovation, however, seems to have developed differently in the Tartessian superlative ending, as
witnessed by uarbaan (J.1.2, J.3.1, J.4.1) ‘the highest one’. Koch, “Paradigm Shift?” 289
41
Ian Maddieson, Patterns of Sounds (Cambridge, Cambridge University Press, 1984) 35-37
is fundamental, as the only basis Koch seems to have to hypothesize that Celtic is Iberianized
Indo-European is the loss of the Proto-Indo-European *p phoneme.42

Furthermore, reflexes of Proto-Indo-European *p do occur in Celtic. A case has


recently been made for the late existence of a reflex of in the form of /ϕ/ in Continental
Celtic.43 Schrijver has even attempted to establish h as a possible reflex in the Old Irish
Würzburg glosses.44 Loss of /p/ is therefore insignificant, and it may even be
counterproductive to the case for Iberia as the Celtic homeland, as the chronology of loss of
PIE *p would not make sense were it to happen as a result of contact with Iberian.

Lusitanian and kw > p

The development of Lusitanian kw > p does not make sense at first glance, as both Celtiberian
and Tartessian are considered Q-Celtic languages lacking this innovation. This may be posited
as an argument to separate Lusitanian from Celtic. P-Celtic languages sharing this innovation
are situated further northwards on the other side of the Pyrenees. It should be noted, however,
that this innovation seems to be an areal innovation between the phylogenetically sister
languages Gaulish and Brittonic. Lusitanian would seem to be an unlikely candidate to
participate in this areal innovation since Celtiberian does not do so, but there is some evidence
for P-Celtic tribal names such as the Pelendones in Celtiberia. Admitting Lusitania to this
Sprachbund would weaken the force of *kw > p in separating Lusitanian from Celtic.45

Significance of the -ko suffix

The -ko- suffix is clearly a descendant of the Proto-Indo-European suffix *-ikos, which has
cognates inter alia in Latin -icus, Greek – ός, and Proto-Germanic *-gaz.46 It is clear from
this reconstruction that the –iko- variant is the most conservative variant, while -aico- and –
aco- may be explained as innovations. Variation between these two (and possibly their
divergence in meaning depending on the preceding vowel) is probably the result of the

42
Joseph F. Eska, review of John T. Koch, Barry W. Cunliffe (ed.), Celtic from the West 2: Rethinking the
Bronze Age and the Arrival of Indo-European in Atlantic Europe. Celtic studies publications, 16. (Oxford;
Oakville, CT, Oxbow Books, 2013) Bryn Mawr Classical Review, accessed January 13, 2014
http://www.bmcreview.org/2013/12/20131235.html.
43
Joseph F. Eska, ‘In defense of Celtic /φ/’, in Adam I. Cooper, Jeremy Rau, and Michael Weiss
(eds.), Multi nominis grammaticus: Studies in Classical and Indo-European linguistics in honor of Alan J.
Nussbaum on the occasion of his sixty-fifth birthday. (Ann Arbor, Beachstave Press, 2013) 32–43.
44
Peter Schrijver, ‘On the nature and origin of word-initial h in the Würzburg Glosses’ Ériu 48 (1997) 205-
27
45
Wodtko, “Lusitanian” 360
46
Calvert Watkins (ed.) The American Heritage Dictionary of Indo-European Roots, 2nd Edition. (Boston,
Houghton Mifflin, 2002) 36
suggested intense two-way language contact between Lusitanian and Celtiberian. Inferring a
close genetic relationship between these languages on the basis of the similar use they make
of this derivation is not possible, however, because the same suffix is also used to derive
adjectives from nouns in other Indo-European languages.

From Indo-European to Celtic

The Iberian hypothesis requires a date for Proto-Celtic which is significantly earlier than the
orthodox view. Furthermore, Iberia is further from the most popularly supposed Indo-
European homelands than Central Europe. This raises the question about how the chronology
fits in with current theories about the relationship among the various Indo-European
languages and the spread of Indo-European across Europe.

Graham Isaac is of the opinion that the chronologies do not add up. Celtic shows,
Isaac argues, a distinctive set of morphological and phonological innovations. Celtic
phonology differs very strikingly from other western Indo-European languages in that the
plain voiced stops and the aspirated voiced stops merged. This development is shared with
Baltic, Slavic, Albanian, and Iranian (but not Indic). A grammatical innovation shared with
Greek, Slavic, Indo-Iranian and Phrygian, but not Italic and Germanic (among others) is the
introduction of relative pronoun *ios, *ieh2, *iod. Similarly, formation of the future tense with
the suffix *-sie-/*-sio- is found in Greek, Baltic, Slavic and Celtic. All of these points argue
for a late Indo-European, and easterly, Sprachbund. If we are to accept that the merger of
voiced stops and voiced aspirate stops in Iranian is connected to the same development in
Celtic, a date between the Indic-Iranian split and the first Avestan texts (ca. 1000 BC) needs
to be posited as a terminus ante quem for Celtic migration westwards.47

Koch’s reply to this criticism by Isaac is that Celtic did not spread to Iberia by means
of migration over land, but rather that Celts were a seafaring people, much like the Greeks.
The above-mentioned developments would have to have occurred some centuries before
1650 BC, at which time Greek was already a fully-fledged written language, and no longer in
contact with Indo-Iranian or Balto-Slavic. Koch’s hypothesis is therefore not a priori
incompatible with Isaac’s chronology. Cunliffe’s proposal, however, most likely is.48

47
Graham Isaac, “The Origins of the Celtic Languages: Language Spread from East to West” in Celtic from
the West: Alternative Perspectives from Archaeology, Genetics, Language and Literature, ed. John T.
Koch and Barry Cunliffe ( Oxford; Oakville, CT, Books, 2010) 158-162
48
Koch, “Paradigm Shift?” 191-192
Conclusion

While the Celticity of Tartessian is itself far from proven, I propose that stating that
Tartessian is (para-)Celtic is a methodologically sound starting point. On the one hand,
Tartessian is demonstrably Indo-European, as witnessed by the endings in narketi and
narkenti in many inscriptions. We also have good reason to postulate a Celtic elite
dominance over Tartessos on the basis of king Arganthonios’s name. On the other hand, we
have been able to distinguish a Celtic and a Lusitanian branch of Indo-European migrations
into pre-Roman Iberia so far, and as long as there is no definitive proof of the precise
linguistic affiliation of Lusitanian, we would be making fewer assumptions by supposing that
it is related to its neighbours Tartessian and Celtiberian. This modus operandi is further
justified in view of the distance of Iberia from any putative Indo-European homeland, making
a theory postulating fewer Indo-European intrusions into Iberia preferable to one postulating a
separate incursion for every separate language attested.

Even if Tartessian is not itself Celtic, it is necessary to assume a variety of Celtic


languages other than Celtiberian to account for the date and form of Celtic elements found in
Tartessian. An example is the dative plurals in –bo instead of –bos. If we are indeed to accept
this dative plural in bo as Celtic, this also makes their distribution problematic and difficult to
reconcile with the traditional siting of the Celtic homeland. A Central European account
modelling the spread of Celtic would not be expected to have the dative plurals in -bo found
in languages in the core area (Gaulish, Lepontic) as well as in peripheral areas (Tartessian,
Celtiberian). A maritime model would make more sense for such a distribution. One might
then expect to find one form (-bo) in coastal areas and another (-bos) in more inland areas.

Nevertheless, the exact phonological reconstruction required to demonstrate the


Celticity of Tartessian is rather unclear. The case for the initially proposed development of
Proto-Indo-European *ṛ and *ḷ as *ri and *li before a plosive has not been conclusively
demonstrated owing to a lack of evidence one way or another.

Using Proto-Indo-European *p has not proved to be a reliable diagnostic as a means


for establishing Celticity for a particular language due to the relatively late date of its loss.
The lack of /p/ in Iberian has been cited as being conducive to Celtic loss of *p. This
reasoning is problematic, however, since loss of /p/ is a relatively unremarkable development
in a language’s phonological inventory. Moreover, no further evidence for an Iberian
substrate has been suggested for Celtic.
The development of *kw > p has been cited for Lusitanian, as it would make little
sense geographically for this phenomenon to be located between the Hispanic Q-Celtic
languages. In the end, however, this would seem to be best explained as an areal feature.

The development of Proto-Indo-European gw > b seems to be the most promising


feature as a diagnostic for Celtic. For Tartessian, it relies on both the correct interpretation of
word boundaries and the correct interpretation of words such as baane ‘woman’. In
Lusitanian, the divine name Bandue is thought to contain the same development. Naturally, it
would be preferable to have examples that are more convincing to demonstrate this
development, but perhaps phonological reconstruction on the basis of such scarcely attested
languages is also best described as ‘making the best use of bad data’.49

An alternative to finding sound laws would be to look for morphological parallels, as


such elements are less likely to be borrowed between languages and matching morphological
innovations may be less likely to be coincidence than phonological innovations. The example
of the descendants of *-ikos, however, proves again that such examples need not only be
present in all the attested languages; they must also show a unique innovation from Proto-
Indo-European.

The Iberian hypothesis for the Celtic homeland pushes back the date of Proto-Celtic
several hundred years. Since Celtic shares a series of phonological and grammatical
developments with several other Indo-European languages, the ultimate date of Celtic leaving
a Proto-Indo-European homeland needs to be pushed forward by an indeterminate amount of
time. Furthermore, Iberia is further away from any hypothesized Indo-European homeland
than the traditionally posited Celtic homeland of Central Europe. The solution to this
complication may lie in reimagining the ancient Celts as a seafaring people.

All in all, the hypothesis that the Celtic language developed in Iberia is supported by a
compelling narrative and has an advantage in that it merges the several strands of Indo-
European found in Iberia into a single ‘invasion’. Despite all this, the current evidence for this
model is scarce, and anyone endeavouring to find more evidence must, one way or another,
elucidate the problems associated with interpretation of the Tartessian inscriptions and the
fact that Lusitanian shows a clear Celtic superstrate.

49
Labov used this phrase for historical sociolinguistics initially, but it may be equally applicable to all
branches of historical linguistics. (in William Labov, Principles of Linguistic Change. Volume 1: Internal
Factors. (London; New York, Basil Blackwell 1994) 11)
Bibliography
Cunliffe, Barry and Koch, John T. “Introduction” in John T. Koch and Barry Cunliffe (eds.),
Celtic from the West: Alternative Perspectives from Archaeology, Genetics, Language
and Literature, 13-38. Oxford; Oakville, CT: Oxbow Books, 2010
Eska, Joseph F. ‘In defense of Celtic /φ/’, in Adam I. Cooper, Jeremy Rau, and Michael Weiss
(eds.), Multi nominis grammaticus: Studies in Classical and Indo-European linguistics
in honor of Alan J. Nussbaum on the occasion of his sixty-fifth birthday, 32–43. Ann
Arbor: Beachstave Press, 2013
Eska, Joseph F. review of John T. Koch, Barry W. Cunliffe (ed.), Celtic from the West 2:
Rethinking the Bronze Age and the Arrival of Indo-European in Atlantic Europe.
Celtic studies publications, 16. Oxford; Oakville, CT: Oxbow Books, 2013, Bryn
Mawr Classical Review, http://www.bmcreview.org/2013/12/20131235.html.
Isaac, Graham, “The Origins of the Celtic Languages: Language Spread from East to West” in
John T. Koch and Barry Cunliffe (eds.), Celtic from the West: Alternative Perspectives
from Archaeology, Genetics, Language and Literature, 153-168. Oxford; Oakville,
CT: Oxbow Books, 2010
Jordán Cólera, Carlos “Celtiberian” in E-Keltoi Volume 6: The Celts in the Iberian Peninsula
(2007)
Jordán Cólera, Carlos Celtibérico. Zaragoza: Ediciones del Departamento de Ciencias de la
Antigüedad, 2004
Koch, John T. “Paradigm Shift? Interpreting Tartessian as Celtic” in John T. Koch and Barry
Cunliffe (eds.), Celtic from the West: Alternative Perspectives from Archaeology,
Genetics, Language and Literature, 185-302. Oxford; Oakville, CT: Oxbow Books,
2010
Koch, John T. Tartessian 2: The Inscription of Mesas do Castelinho ro and the Verbal
Complex. Preliminaries to Historical Phonology, Oxford; Oakville, CT: Oxbow
Books, 2011
Labov, William, Principles of Linguistic Change. Volume 1: Internal Factors. London; New
York: Basil Blackwell 1994
Lorrio, Alberto J. and Ruiz Zapatero, G. “The Celts in Iberia: An Overview” in E-Keltoi
Volume 6: The Celts in the Iberian Peninsula (2005): 167-254
Maddieson, Ian, Patterns of Sounds, Cambridge: Cambridge University Press, 1984
Matasović, Ranko, Etymological Dictionary of Proto-Celtic. Leiden; Boston: Brill, 2009
Prósper, Blanca Ma, lenguas y religiones prerromanas del occidente de la Península Ibérica,
Salamanca: Ediciones Universidad de Salamanca, 2002
Schrijver, P.C.H. ‘On the nature and origin of word-initial h in the Würzburg Glosses’ Ériu 48
(1997): 205-27
Untermann, Jürgen and. Wodtko, Dagmar S. Monumenta Linguarum Hispanicarum 4. Die
tartessischen, keltiberischen und lusitanischen Inschriften, Wiesbaden: Ludwig
Reichert, 1997
Watkins, Calvert, (ed.) The American Heritage Dictionary of Indo-European Roots, 2nd
Edition. Boston: Houghton Mifflin 2002
Villar, Francisco, ‘Indo-Européens et Pre-Indo-Européens dans la Péninsule Ibérique’ in
When Worlds Collide : Indo-Europeans and Pre-Indo-Europeans, in T.L. Markey and
J.A.C. Greppin (eds.), 363-94. Ann Arbor: Karoma, 1990
Villar, Francisco, “The Celtic Language of the Iberian Peninsula” in Studies in Baltic and
Indo-European Linguistics in Honor of William R. Schmalstieg Ed. Philip Baldi and
Pietro U. Dini, 243–274 Amsterdam: John Benjamins, 2004
Wodtko, Dagmar S. Ancillary Study: The Problem of Lusitanian” in John T. Koch and Barry
Cunliffe (eds.), Celtic from the West: Alternative Perspectives from Archaeology,
Genetics, Language and Literature, 335-368. Oxford; Oakville, CT: Oxbow Books,
2010
Wodtko, Dagmar S. An outline of Celtiberian grammar, Freidok: Albert-Ludwigs-Universität
Freiburg, 2003,
http://www.freidok.uni-freiburg.de/volltexte/747/pdf/Celtiberian.pdf

You might also like