The Question of Italo-Celtic Unity

Anthony D. Yates

The Italo-Celtic hypothesis, an Indo-European subgroup uniting Italic and Celtic

into a single entity, has, since its inception in 1861 (Lottner Kuhns Beiträge 2.309 ff.),

sufficiently intrigued generations of scholars as to gain a kind of cyclical immortality.

This sort of immortality is not without death; on the contrary, the theory has perished

many times, 1 but has always been resurrected—including, notably, by Cowgill 1970 from

Watkins 1966. 2 The principal appeal of this and other theories of subgrouping lies in their

contribution toward the resolution of a question fundamental in Indo-European studies, as

framed by Watkins: “wie es eigentlich gewesen?” (1966: 29). An intermediate Italo-

Celtic subgroup, existing in the vast temporal grey space between Proto-Indo-European

and its relevant end-points, i.e., the daughter languages of the separate Italic and Celtic

subgroups, provides valuable insight as to the process.

The hypothesis itself originated to explain several shared innovations between the

two branches. To these, the principal criteria for linguistic subgrouping, Watkins would

add shared retentions and divergences as equally valid measures of linguistic

verisimilitude—and rightly. 3 The principle evidence for Italo-Celtic unity is compiled by

Meillet in his seminal work Les dialectes indo-europeéns (1908: 51-56), an essentially

1
Watkins 1966: 29 n.1 lists at least twelve such studies affirming or rejecting the hypothesis.
2
Cowgill employs a memorably grisly metaphor: “[I]n reopening the grave just closed by Watkins I have
found neither an empty coffin, nor a living adult prematurely buried, but a stillborn infant.” (And yet we
wonder why linguists may not always be invited to parties.)
3
Although he draws very different conclusions from Watkins, Cowgill 1970:193 affirms the significance of
these data-points.
exhaustive list of features common to Celtic and Italic alone. This list of seven items, as it

provides the fundamental shape for later discourse, is reproduced here in an Appendix. 4

Items 2), 5b), 7), and 8) are dismissed in Watkins 1966. In the first, he cites a 1956

study of J. Kurylowicz, 5 which shows that this development of IE. *r̥ , *l̥ is not unique to

Italo-Celtic, but is common to a larger group including Greek and Armenian. Similarly,

Watkins’ own earlier work 6 has proven the shared property 5b) to be a “historical

accident:” the origins of the Celtic subjunctive in -s- is a Celtic aorist indicative, whereas

in Latin, an -s- future). Item 7) he rejects on the grounds of incomparability of structural

contexts; its development he regards as important only insofar as it represents a broader

shared “negative” innovation 7 —viz. a limiting of the productivity of deverbative nouns.

The final item is likewise disallowed, citing, inter alia, the general weakness of such

evidence, and the absence of Celtic cognates for Latin ab, ante, apud, circum, ob, per,

and prae.

The striking innovation of 3) remains unclear. It is not exclusively Italo-Celtic;

Messapic too shows stems in -o with genitive singular in –ihi, supposed to be -ī. Watkins

1966 notes significant variation in these forms in the various Indo-European languages,

and so suggests the possibility that there was no original Indo-European form; innovation

is thus to be expected, and the shared isogloss in Italic and Celtic may be due to contact.

Items 1), 4), 5a), and 6) have been a source of significant disagreement. As to 4),

4), Watkins is skeptical, citing the remarks of Kurylowicz (op. cit); in contrast, Cowgill

4
The Appendix provides easy reference (sc. detachable) to the individual items, which will be cited by
number throughout this work.
5
L’apophonie en indo-europeén 166 ff. (Wroclaw, 1956)
6
Indo-European origins of the Celtic verb. I. The sigmatic aorist (Dublin, 1962)
7
Watkins believes that “negative” innovations, properties of the proto-language that have been restricted or
eliminated in its daughter languages, play an important role in subgrouping alongside the more commonly
cited “positive” innovations.

2
emphatically concludes that the Celtic and Italic use of “-r as a voice marker pure and

simple IS an innovation,” comparing Hittite and Tocharian which use it primarily as a

tense marker (1970: 221). Watkins’ disputations of 1) and 5a) can essentially, with

respect to the views articulated by Cowgill, be reduced to problems of relative

chronology. The debate on 1) focuses on Latin quercus, cognate with Celtic toponym

Ἑρκύνια. The latter indicates that in Celtic the assimilation of p…kw to kw … kw occurs

after the change of kwu to ku, while the former, the opposite chronology. While Watkins

believes that this is sufficient evidence for independent development,8 Cowgill 1970: 191

doubts its significance, certainly to the extent that it rules out a Italo-Celtic subgroup.

Similarly for 5a), the major differences in the modal systems of Celtic and Italic

identified by Watkins 1966: 41-3 are, in the view of Cowgill, the result of secondary

divergence postdating the creation of -ā- subjunctive ending. Finally, with regard to 6),

Cowgill 1970 brilliantly demonstrates that the foremost problem advanced by Watkins

with Meillet’s original theory—viz. the existence of two separate superlative suffixes

*samo and *isamo—is no problem at all: the suffixes can be reconciled to a single

*-ism̥ mo-; hence, they are absolutely and entirely of their own kind, a unique innovation

common to Celtic and Italic.

Based on a consideration of this evidence, Watkins 1966 firmly rejects the notion

of Italo-Celtic unity, concluding: “The only common language from which both Italic and

Celtic can be derived is Indo-European itself” (43-44). Yet the same evidence has led

Cowgill 1970 not just to posit unity, but to forward a plausible migratory theory to

account for it. His theoretical model assumes an extreme western Italo-Celtic group

8
Watkins 1966: 34 draws additional support for this conclusion from a proposed derivation of déac from
*dwei-penkw-. Cowgill 1970: 191-2 n.1 disputes this derivation with good cause, favoring a derivation from
*dekan-os, from which the OIr. deec by a metasthesized *deankos.

3
speaking a very late form of Proto-Indo-European, including only very basic

developments (such as laryngeal loss) and its distinct set of shared innovations. Their

contact was terminated by an influx of Germanic peoples, setting the stage for a

millennium or so of independent development (222-3). While a reconciliation of

Watkins’ conclusion with this account may be no easy task, it is certainly not the

proverbial case of a square peg and a round hole.

While the efforts of Cowgill 1970 have been supplemented and reinforced by,

inter alii, Kortlandt 2007: 25 ff., which presents further evidence for an early period of

common development followed by a much longer divergence before the first extant texts

(25 ff.), a general acceptance of a Italo-Celtic unity is, however, far from a reality. The

irresolution of 3)—long held to be important evidence for unity, as well as the

tremendous differences these systems manifest by the period of the earliest texts present

for many scholars a formidable obstacle to effective subgrouping. Moreover, one of many

potential fissures may be found in the developing controversy over Italic unity, a question

which would seem to presuppose an Italo-Celtic unity. 9 Nevertheless, barring significant

new evidence, Cowgill’s irrefutable demonstration of the unique, shared innovation of

the superlative suffix has likely guaranteed to the Italo-Celtic hypothesis a future, if not

one of universal acceptance, then one outside any graveyards.

9
Watkins 1970 puts aside the question, as his rejection of Italo-Celtic unity renders it irrelevant for his
purpose, and because the issue was treated separately at the same conference and appears in the same
publication (Beeler, M. “The Interrelationships Within Italic.” op. cit. 51-58). Needless to say, much has
been discovered, since Meillet—perhaps flippantly—discarded the possibility of a separation in Italic
(“These small common features establish the validity of a period of Italic unity” [1908: 50]); thus the
question is, at least, worthy of consideration in any discussion of Italo-Celtic unity.

4
 References

Cowgill, Warren. 1970. “Italic and Celtic Superlatives and the Dialects of Indo-
European.” In: The Collected Writings of Warren Cowgill. Jared Klein, ed. New
York: Beech Stave Press, 2006.

Kortlandt, Frederik. Italo-Celtic Origins and Prehistoric Development of the Irish

Meillet, Antoine. 1929. Ancient Indo-European Dialects. transl. Samuel N. Rosenberg.

Watkins, Calvert. 1966. “Italo-Celtic Revisited”. In: Ancient Indo-European Dialects: 29-
50. Birnbaum, Henrik; Puhvel, Jaan eds. Berkeley: University of California Press,
1966.

5
 Appendix

1) p…kw > kw… kw (Lat. quinque, Ir. cōic, Wel. pimp < IE. *penkwe; cf. Gk. πέντε, Skt.
páñca)
2) The separate treatment of IE. *◦r, *◦l (> ar, al) and *r̥ , *l̥ (> Lat. or, ol, Celt. ri, li)
3) Stems in -o with genitive singular in -ī (Lat. virī = OIr. fir, Ogam Ir. maqi)
4) Passive in -r 10
5) Subjunctive endings in a) -ā- and b) -s- (Lat. feram, OIr. bera; Lat. dīxim, faxim, OIr.
tīasu, tēis)
6) The superlative suffix *-samo- (Lat. maximus, O. nessimas, Umbr. nesimei, Wel.
nesaf)
7) The suffix –tiō/n- 11
8) Identical vocabulary, esp. prepositions and preverbs (Lat. dē = OIr. dī, Bryth. di; Lat.
cum=Ir. com)

10
It only later became known to Meillet that this was a shared retention, not an innovation (Watkins
1966:39)
11
Meillet calls this suffix *-tei- “supplemented by a nasal infix