Published online by Cambridge University Press: 22 April 2026
The dialect area with which this study deals is the Roman imperial administrative district of Narbonensis, the southeastern part of Roman Gaul—in essence, the modern Provence. A more exact delimitation of this region, together with a map, may be found in DAG pp. 59-62 and 14. All the extant evidence of the Gaulish language has been collected and classified by What-mough in DAG; with one or two exceptions from CIL 12, the material for this study has been drawn in its entirety from the section on Narbonensis, DAG 20-83.
1 Abbreviations for onomastic categories: PN—personal name, DN—divine name, LN—local name (plurals PNN etc.); for regions of Gaul: Narb.—Narbonensis, Aquit.—Aquitania, Lugd.—Lugdunensis, Belg.—Belgica, G. Sup.—Germania Superior, A. Dec.—Agri Decumates; for source: DA G—Joshua Whatmough, The dialects of ancient Gaul (University Microfilms; Ann Arbor, 1949-51). References to DAG are to sections.
[Gaulish forms preserved in Greek spelling are cited in boldface transliteration. Greek letters in forms otherwise written in the Latin alphabet are shown similarly. Note th for theta, ph for phi, kh for chi.—Editor.]
I wish to express my gratitude to Joshua Whatmough, Roman Jakobson, and Lawrence G. Jones for comments and helpful criticisms, and to acknowledge my debt to Whatmough's DAG, without which this study would not have been possible. Thanks are due also to the Social Science Research Council for a grant which enabled me to carry on research in Gaulish during 1953.
2 For the problems in analyzing such languages phonemically, see Horace G. Lunt, On Old Church Slavonic phonemes, Word 8.311-29 (1952).
3 Some doubt exists about the position between initial consonant and vowel, as in giamo-and suadu-.
4 Still, the spelling is ambiguous, as ou represents [ow], phonemically /ou/, in certain other forms before a syllabic (e.g. Rouicus, roouika, Douiccorix) and always before a non-syllabic. This probably accounts for the infrequency of ou for [uw].
5 See Whatmough, Die Sprache 1.126-7 (1949); HSCP 60.175-85 (1951).
6 It is not known under what conditions such clustering takes place. Whatmough suggests substratum influence (Die Sprache 1.126), but the wide geographical distribution of this feature is against it. Jakobson has pointed out to me the existence in Slavic of a similar divergence in the development of clusters: *lowjǫ (lo-wj-ǫ) > lovlju beside *darowjǫ (dar-ow-jǫ) > daruju.
7 Even before the Romans came, /ūi/ had a tendency to lose the second member. Beside eight occurrences of /ūi/ at this time we have two certain and one possible case of /ū/ with the same morphemic function in Narbonensis.
8 Cf. Kenneth Jackson, Language and history in early Britain 306 (Cambridge, Mass., 1953).
9 See esp. Tenney Frank, Tau Gallicum, Vergil, Catalepton II.4, AJP 56.254-7 (1935). For a discussion and further bibliography see DAG 79 s.v. tau.
10 Om runene og runenavnenes oprindelse, NT 1.85-188, esp. 104-13 (1928). Jackson (op.cit. 531) says from the historical point of view that it is better to take the source of Brythonic ss, st as ts, although ‘the exact position of Đ in the scheme is hardly clear.‘
11 See the author's review of Jackson, op.cit., Lg. 30.513-8 (1954).
12 Cf. the situation in early Germanic, e.g. OHG hypocoristic Sicco beside Sige-rīch, Sig-fridus. See Brugmann, Gdr. 2.1.34; Martinet, La gémination consonantique d'origine expressive dans les langues germaniques 194 ff. and passim (Copenhague, 1937). It is not always easy to distinguish such cases from pairs of different lexemes with a fortuitous resemblance, e.g. brogi- ‘land’ brokko- ‘badger’.
13 The history of this pair is more complex. The former is actually Lat. carpentum ‘chariot’, borrowed from Keltic at an earlier period, in Italy; the other is the true Keltic form. I give the pair nevertheless as illustrating the confusion between p and b in the transcription of Keltic.
14 J. Pokorny, Keltologie 135 (Bern, 1953). For detailed arguments see his study, Vox rom. 254-67 (1951).