Wednesday, September 28, 2011


Barry W. Cunliffe, John T. Koch (ed.), Celtic from the West: Alternative Perspectives from Archaeology, Genetics, Language, and Literature. Celtic Studies Publications 15. Oxford/ Oakville, CT: Oxbow Books, 2010. Pp. vii, 384. ISBN 9781842174104. $80.00.

Reviewed by Jürgen Zeidler, University of Trier (

Version at BMCR home site

Table of Contents

This rehabilitation of the "Atlantic hypothesis", that the Celtic languages emerged in the Atlantic Zone during the Bronze Age, pursues previous efforts undertaken by the authors (see p. 1, n. 1). It is a comprehensive and fully illustrated volume with provocative considerations, which are at the cutting edge of research.

The book contains an introduction by the editors (p. 1–10) and eleven individual studies divided into three parts, covering archaeology, genetics and language/literature. Koch's study of Tartessian as Celtic (p. 185–301) comprises almost one third of the text of the volume. Due to space limitations, only a few topics can be commented on here.

In the introduction, the term "Celtic" is defined to mean "… a proven affiliation with the Celtic languages or (for non- linguistic evidence) a demonstrable close connection with them" (p. 2). This is both unambiguous and applicable to all geographical areas of Celtic Studies and is preferable to that of the "Vienna school", which leaves the definition to each individual discipline. But the ultimate consequence of the authors' approach is to completely resign from the use of "guide fossils" of "Celtic" material culture.

In line with the last statement, Raimund Karl (p. 39–64) disentangles the historical and modern constructs of the "Celts", which consist of elements of various origins. Besides the archaeological La Tène and linguistic Celtic origin, he identifies a "historical druidic origin" in the British Isles. However, this is in itself a disputable construct because it is not clear whether Caesar's reference to Britain (De bello Gallico 6. 13–14) has any bearing on the origin of druidism.1

Barry Cunliffe (p. 13–38) argues that the geographical setting of the Proto-Celtic language in the Atlantic Bronze Age, c. 2000–750 BC, is in better agreement with the archaeological data than the traditional opinion. One major argument is seen in the large-scale exchange of goods and ideas in the Bronze Age before the fragmentation and regionalization in the transition to the Iron Age. Another important argument is said to be the lack of archaeological evidence for any considerable immigration into the British Isles. Finally, he asks (p. 34) whether Celtic could have developed between 5000 and 3000 BC in the Atlantic Zone. However, this would be an unusually early date from the linguistic point of view. Most scholars accept that late Proto-Indo-European was still a coherent group in the fourth (and even third) millenium BC. Moreover, language spread is not necessarily associated with mass dissemination of material culture, and it is a proven fact that Hallstatt artifacts did spread into the British Isles. Thus it is difficult if not impossible to draw far-reaching conclusions from objects to languages.

In support of the Atlantic hypothesis, Stephen Oppenheimer's alleged chalcolithic "gene-flux" from the Balkans towards the British Isles is brought up as an argument. But Oppenheimer himself (p. 146) is cautious enough to stress that the contribution of genetics "may be an impossible task since genes do not carry ethnic or linguistic labels". And post-neolithic migration resulted in "a minority intrusion to the gene pool".2

Ellen C. Røyrvik (p. 83–106), Brian P. McEvoy and Daniel G. Bradley (p. 107–120) are also reserved regarding conclusions from our present knowledge of the genetics of Britain and Ireland to the linguistic situation. They agree that "population genetics should be able to make a considerable contribution towards the elucidation of Celtic … prehistory" (p. 102), but they also agree that "we are still restricted to examining but a fraction of the human genome's diversity" (p. 118) and are hence unable to reconstruct prehistory from the genetic diversity. Inconclusive as genetic studies presently are, they look more promising for the future. The genetics of facial features (p. 87), the correlation of surnames and Y chromosomes (p. 114–117), to take just these two examples, are all interesting and encouraging.

The philological part of the book begins with a contribution by Graham R. Isaac (p. 153–167), who advocates the traditional viewpoint of language spread from east to west. In particular, he underlines features of Celtic "shared with specifically Iranian" (p. 163). David N. Parsons (p. 169–184) poses the difficult question whether distributional patterns of Celtic place-names reflect "forces at work many centuries after the original Celticization" (p. 182), even "something inherent in the Roman Empire" (p. 183). In two ancillary studies, Philip M. Freeman (p. 303–334) assembles ancient references to Tartessos from Greek, Latin, Assyrian and Hebrew sources (the two latter without the original texts). And Dagmar S. Wodtko (p. 335–367) dwells on the problem of Lusitanian. Her thorough and cautious analysis concludes with the remark that "the assessment of the linguistically Celtic and non-Celtic features in western Spain, Galicia and Portugal depends crucially on the interpretation of the evidence" (p. 362), specifically with regard to etymologies, historical phonology and language affiliation.

Pursuing previous studies, John T. Koch (p. 185–301) argues that the so-called Tartessian inscriptions reflect a Celtic language. In his analysis, he includes new inscriptions discussed by Amílcar Guerra (p. 65–79), in particular from Mesas do Castelinho (meanwhile also in Tartessian 2. The inscription of Mesas do Castelinho and the verbal complex. Oxford: Oxbow, 2011). Apart from these inscriptions, the presence of Celtic speakers by the time of Herodotus (c. 484–425 BC) or earlier is suggested by personal and tribal names such as Arganthonios (*arganto- 'silver') and Kunetes (*kun-et- 'hound (warrior)'). Whether this has a bearing on the Tartessian inscriptions of the eighth to sixth centuries BC, however, remains problematic.

The inscriptions are written in a variant of the Palaeo-Hispanic scripts (p. 203–208), which are semisyllabic. Alphabetic signs for vowels (a, e, i, o, u), resonants (m, n, r, ŕ, l) and sibilants (s, ś) are distinguished from syllabic signs for stops plus vowel (Pa, Pe, Pi, Po, Pu etc.). Stops are only divided according to their place of articulation (P: labial, T: dental, K: tectal), not to sonority (voiced, voiceless). No word dividers are used. A typical feature of Tartessian is the use of extra vowel signs usually accompanying syllabic signs with the same vowel, e.g. Po+o. Thus the Tartessian script resembles an intermediary between semisyllabic and alphabetic writing. Some cases of disagreement occur, though, e.g. Ko.o.ŕ.Pe.o (J.53.1)—interestingly always Ce +V—and rarely no extra vowel is written after a stop, e.g. Tu.n.?i.i.te.s.Pa.a.n (J.53.1).

It should be clarified from the outset that a system like this—hardly suitable for the denotation of an Indo-European language as it is—leaves ample room for interpretation.

Scanning through the suggested translations, one cannot help feeling puzzled by the results achieved by the author. One wonders automatically what message these texts intend to convey. Moreover, epigraphic and literary styles seem to change arbitrarily. Elements of funeral inscriptions appear to be combined with those of dedications and elegies, e.g. "Invoking the divine Lugoves of the Neri (tribe), this funerary monument for a noble … 'Celt/Gaul'. Invoking all the heroes (Eśkingolī group) [the necropolis] has received, [the grave] of Ta[χ]seovonos" (J.1.1, p. 211). What are the invoked groups expected to do for the deceased or the grave? Obscurities like this arouse one's suspicion.

A closer look at orthography and phonetics reveals a number of inconsistencies, even if Koch's linguistic analysis is followed as far as possible. Thus, reconstructed Proto-Celtic (short) */e/ is taken to be variously written e, i and even ii, cf. i.ś */eχs-/ 'out of' (J.1.1), n.i.i.r.a.Po.o */nerabo/ 'belonging to the Neri' (J.1.1 – sic! for -abo is the feminine dative plural ending, thus the nominative plural should be */Nerās/).3 An ad hoc assumption of a phonetic change is sometimes visible.4 And the stem vowel of the o-declension fluctuates between o and a.5

As regards morphology, a strange combination of archaic and unexpected young traits can be observed. One example may serve as an illustration: the use of To.o */do/ 'to' (J.1.1) to reinforce the dative would be rather unusual for an ancient Celtic language.6

To sum up, Koch's analysis reflects the author's superior scholarship, but is not really convincing. The reader is left with a number of inconsistencies, in form and content, ad hoc solutions and divergencies from the results of the other Hispano-Celtic sources. Nevertheless, it is a strong vote for a Celtic solution to the problem of Tartessian, and future research will not be able to avoid this approach. As in the case of Lusitanian, it may very well be a hybrid language with a non-Celtic matrix and extensive Celtic loanwords (as previously assumed by Francisco Villar) or vice versa.

In the end, the hypothesis of an Atlantic origin of the Celtic languages remains a problematic issue. Genetics is not yet (and may never be) able to give any clues beside support for the accepted view that no mass immigration into the British Isles ever happened in the second half of the first millenium BC. Archaeology cannot provide "guide fossils" for communities that are primarily defined by linguistic criteria. Trade connections and "elite networks" affect only small fractions of society with little impact on the choice of language. As to the linguistic evidence, Parsons's toponomastic study questions the time-depth and thus the significance of the Celtic nomenclature (although more material has to be analysed to reach reliable conclusions). Written sources do show a Celtic element in the Iberian Peninsula as early as the sixth or fifth century BC, but this is late enough to allow for an Atlantic as well as a Central European origin of Proto-Celtic. The results from historical linguistics are so far not decisive. Correspondences with various Indo- European languages do not prove more than subsequent phases of mutual influence. It is hardly possible to determine the space and time of these contacts.

One could assume that loanword, substrate and hydronym analysis, together with advanced methods of "linguistic palaeontology"7 rather than genetics and archaeology, will contribute to a better understanding of the linguistic development of Celtic.


1.   A. Hofeneder. Die Religion der Kelten in den antiken literarischen Zeugnissen. vol. 1. Vienna: Austrian Academy, 2005: 195–196.
2.   He thinks of the dissemination of the haplogroups J2e-M12 and E3b1a2 of the Y chromosome as a hint at mining as a motive for long-distance migration in the Chalcolithic and early Bronze Age. But it may be asked whether groups of miners would really leave significant traces in the gene-pool of the native inhabitants.
3.   The following examples are completely unconvincing: */ei/ written eai (Te.e.a.i.o.n.a */deiwonā/ 'goddess', J.4.3), */a/ (Koch: */ə/) written eo (r.i.n.o.e.Po.o */rī(g)nabo/ 'to the queens', J.5.1).
4.   In a.r.i.a.r.i.ś.e */ario-rīgi/ 'for Arioriχs' (J.10.1): ś is said to represent palatalized */g'/ before */i/ (n.b.: the dative ending could as well be */ei/ as in Celtiberian) whereas in the similar case of */k/ before */i/, or other consonants, no such change is assumed (e.g. ki.i.e.l.a.o.e */kīlawāi/ 'for Kilawa', J.11.1).
5.   As in the last example a.r.i.a */aria-/ (J.10.1) compared to l.e.Po.o-i.i.r.e */lemo-wirī?/ (Mesas do Castelinho), u.a.r.Po.o-i.i.r */uwarmo-wir-/ (J.22.1) and (o).i.r.a */wira/ 'man' without -s (J.1.2) in contrast to Ti.i.r.To.o.s */tirtos/ 'third' (proper name, J.1.2). Ka.a.k() */kʷākʷ-/ 'all' (J.1.1, n.b.: *Ku.a.Ku- would have been easily possible) remains without an ending altogether.
6.   TO LVGVEI in Peñalba de Villastar (K.3.3), which is quoted in support of this analysis, has been explained in many different ways (D.S. Wodtko. An outline of Celtiberian grammar. Freiburg 2003: 17). A connection with the theonym has been rejected, in favour of an abstract noun 'oath', by F. Beltrán Lloris et al. (Palaeohispanica 5 (2005): 914–930).
7.   Zsolt Simon. "How to find the Proto-Indo-European homeland? A methodological essay". Acta Antiqua Hungarica 48 (2008): 289–303.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.