No CrossRef data available.
Article contents
Celtic Lenition and Western Romance Consonants
Published online by Cambridge University Press: 22 April 2026
Extract
‘Lenition’, in German ‘Lenierung’, a term proposed by Rudolf Thurneysen, has nowadays widely superseded the traditional but ambiguous 'aspiration' as the designation of a set of phenomena characteristic of all insular Celtic languages. The least ambiguous definition of 'lenition' is probably that to be found in Thurneysen's Grammar of Old Irish, a revised and enlarged translation of his Handbuch des Altirischen: 'Lenition ... is the term used to describe a mutation of consonants which normally originated in a reduction of the energy employed in their articulation. It affected not only medial, but also such initial consonants as were closely associated with the preceding word. ... It is earlier than the loss of vowels in final and interior syllables ..., for it presupposes the continued existence of these vowels.' In the ensuing paragraphs, Thurneysen indicates the phonetic contexts in which Irish lenition took place; but since we are here concerned with the phenomenon of lenition in general, rather than with the details in any specific form of insular Celtic, we need not list all the situations in the various Goidelic and Brythonic languages in which lenited consonants occur—the less so since we find in attested linguistic stages traces of an extension of the lenited pronunciation to cases where it certainly did not exist at the time when the actual leniting shift originated. Let it suffice here to say that lenition proper arose when a consonant was surrounded by open articulations, mostly vowels and w, but also, in the case of some consonants, when followed by l, r, n. For the sake of simplicity, in what follows we shall frequently use the phrase 'intervocalic position' in reference to all the contexts in which lenition took place.
Information
- Type
- Research Article
- Information
- Copyright
- Copyright © 1952 Linguistic Society of America
References
1 IF Anz. 9.42.
2 Cf. H. Pedersen, Aspiration i irsk (Leipzig, 1897); J. Vendryes, Grammaire du vieil-irlandais 23 (Paris, 1908).
3 74 (Dublin, 1946).
4 Cf. e.g. Thurneysen 75 on the lenited pronunciation of OIr. former geminated l, r, n immediately following a stop as a result of syncope.
5 The same will apply to Western Romance, where the shift from t to d in patrem > padre will be tacitly assumed to be covered by the statement that intervocalic -t- is shifted to -d-.
6 A concise comparative Celtic grammar 127 (Göttingen, 1932).
7 Breton, though at present spoken on the continent, is of course rightly classified as Insular Celtic.
8 Thurneysen 218.
9 Joseph Vendryes sets apart most carefully the phonetic phenomenon and the morphophonemic process when he refers to lenition and to other similar phenomena, Grammaire du vieil-irlandais 53, as 'certains phénomènes phonétiques dont l'application est antérieure à l'époque des plus anciens textes et qui n'ont d'intérêt en vieil-irlandais que parce qu'ils y ont donné naissance à des alternances morphologiques'. The two aspects of ‘lenition’ are clearly contrasted by D. M. Jones, Actes du 6e Congrès international des linguistes 241 f. (Paris, 1949), and F. Falc'hun, Le système consonantique du Breton 84 f. (Rennes, 1951).
10 Mutation in Gaulish, Lg. 20.223–30 (1944).
11 G. Dottin, La langue gauloise, inscr. no. 59 (Paris, 1920).
12 Carl J. S. Marstrander, NTS 1.116; Leo Weisgerber, Galatische Sprachreste, Natalicium Johannes Geffcken 168 fn. 4 (Heidelberg, 1931).
13 From a theoretical phonemic standpoint, it could be said that even if [p] and [h] remained in complementary distribution, they would have to be reckoned as actualizations of two different phonemes, since their only common relevant feature is their consonantal articulation; cf. Troubetzkoy, Principes de phonologie 37 f. In practice, proof of the phonemic independence is often given by loanwords or expressive creations. Thus, in the traditional vocabulary of Norwegian, [ç] and [k] might seem to stand in complementation —[ç] before front vowels, [k] elsewhere—but the existence of the loanword arkiv and the ‘lockwort’ kiss, both with [k], is a clear indication that we have to posit two phonemes /ç/ and /k/. At the stage of Early Celtic that we are considering here, we may well imagine that [h] soon appeared in utterance initial in loans, interjections, and various more or less marginal forms.
14 Lewis-Pedersen 17.
15 Lewis-Pedersen 18.
16 Lewis-Pedersen 59 ff.
17 See Robert Fowkes, Initial mutation of loanwords in Welsh, Word 5.205 ff.
18 We are here referring mainly to the phenomenon called the spirant mutation of Brythonic, whereby original fortes, either geminated or single, after a consonant appear as spirants (Lewis-Pedersen 122 f.). That this change must be comparatively recent is clear from the fact that the various Brythonic languages do not entirely agree on the conditioning of the change: Welsh has -th- for the -t- of -ltr- and -ntr-, whereas Cornish and Breton have preserved -t- (cf. Lewis-Pedersen 47 f.). Moreover, there is little agreement on the morphological conditioning of this mutation in word-initial position (Lewis-Pedersen 123). All of this contrasts with the universality and regularity of (say) Brythonic lenition of voiceless to voiced stops.
19 As for instance in Welsh rhech ‘fart’ if from 
(Osthoff, ZfcPh. 6.396).
20 See Thurneysen 92 f. This alleged sound shift was first assumed for Germanic, where it finds just as little justification as in Celtic; cf. W. Wissmann, Nomina postverbalia in den altgermanischen Sprachen 160 ff. (Göttingen, 1932), and André Martinet, La gémination consonantique d'origine expressive dans les langues germaniques 84 ff. (Copenhagen, 1937).
21 Cf. A. Graur, Les consonnes géminées en latin 33 f. (Paris, 1929).
22 Cf. Martinet, La gémination consonantique 114 ff.
23 Cf. above, footnote 18.
24 Cf. the situation in Swedish as illustrated by Martinet 20 f., 159.
25 This, of course, presupposes that in the pattern there was no -dd-, or at most very few, which would tend towards -d:- and ultimately -d-, as we shall assume, in what follows, for Goidelic.
26 Thurneysen 126 f.
27 See Alf Sommerfelt's suggestions, Système quantitatif celtique et voyelle additionnelle irlandaise, NTS 5.126 ff.
28 See Thurneysen 90; the view that OIr. non-lenited occlusives were strongly articulated is aptly presented by Sommerfelt, NTS 5.123 ff. But cf. Marstrander's reservations, Okklusiver og substrater, NTS 5.258 ff.
29 Cf. e.g. J. Cantineau, Essai d'une phonologie de l'hébreu biblique. BSL 46.86, 100 ff., 105 f. We need not take a position here as to whether, in a synchronic description of Hebrew, spirants and stops should be considered variants of the same phonemes, which is Cantineau's view, or distinct units (cf. Z. S. Harris, Linguistic structure of Hebrew, JAOS 61.147). Yet we should be tempted to interpret the Massoretic orthographic identification of stops and geminates as an indication that at that time, at least, the ‘geminates’ were nothing but energetically articulated stops, phonemically and perhaps phonetically identical with the stops; then of course spirants would have to be interpreted as independent phonemes.
30 See e.g. Lewis-Pedersen 40: 'IE. q ... became k' in Celtic.'
31 For the details, see Lewis–Pedersen 53 ff.
32 Lewis–Pedersen 55 f.
33 Cf. Lewis–Pedersen 48 ff.; but what is said there about the merger of lenited and non-lenited l, n, r has to be corrected by reference to F. Falc'hun, Le système consonantique du breton 65 ff., from which it is clear that the oppositions have been preserved in Armorican.
34 Thurneysen 85 f.; also 23.
35 See e.g. Martinet, Occlusives and affricates with reference to some problems of Romance phonology, Word 5.120 f.
36 See Haudricourt–Juilland, Essai pour une histoire structurale du phonétisme français 52 ff. (Paris, 1949).
37 We should of course keep in mind that there was a merger in Eearly Celtic.
38 In the rapid survey which follows we shall, as regards Gallo-Romance dialects, rely upon Haudricourt–Juilland 53 ff.
39 The situation in Catalan is a bit confused: we find sòn < *sonnu <somnum by the side of dany < damnum. Bourciez, Eléments de linguistique romane3 313 (Paris, 1930), restricts palatalization of -nn- to Catalan word-final position as in any < annum. Yet in view of canya < cannam, penya < *penna, we shall rather consider, with Meyer-Lübke, Das Katalanische (Heidelberg, 1925), and A. Griera, Gramàtica històrica del Català antic 73 (Barcelona, 1931), that -ny- is the regular reflex of -nn-.
40 Leonese forms are quoted from M. C. Casado Lobato, El habla de la Cabrera Alta (Madrid, 1948). A tilde over the letter denotes palatalization.
41 Cf. Sommerfelt, Système quantitatif 124: ‘comparées aux consonnes du degré faible, les consonnes du degré fort ont été caractérisées par une pression plus grande de la langue ... par une étendue du contact des organes articulateurs plus large ...‘
42 See Haudricourt-Juilland, and, for Piedmontese, J. P. Soffietti, Phonemic analysis of the word in Turinese 38 (New York, 1949). It is most likely that the distinction is further preserved in other sections of northern Italy and Romansh-speaking Switzerland. See Meyer-Lübke, Die Entwicklung von zwischensilbischen n, ZfrPh. 41.555 ff., where however no indication is given of the development of -nn- in the dialects taken into consideration. G. Rohlfs' treatment of the problem, Historische Grammatik der italienischen Sprache 370 f., 394 f. (Bern, 1949), is not quite convincing because he fails to deal with the opposition -n-/-nn- as a whole.
43 G. Rohlfs, Zur Entwicklung von -ll- im Romanischen, Berliner Beiträge zur romanischen Philologie 1.390 ff., assumes a cacuminal stage for which, however, there is no support in Gascon. From an articulatory standpoint, cacuminals and palatals are very different.
44 Rohlfs, Historische Grammatik 364 f.
45 See e.g. Jean Séguy, Le français parlé à Toulouse 19 ff. (Toulouse, 1950).
46 Literature in Haudricourt–Juilland 56 f.; cf. also Jean Séguy.
47 Rohlfs 372.
48 Rohlfs 396.
49 We assume that a in open syllable was first diphthongized to [aḙ] (cf. manu- > main); but if one prefers to assume a > ę, the vowel must all the time have had long quantity, which would keep it distinct from ę in originally checked syllables.
50 Any -īnn- or -ūnn- must have passed to -inn- or -unn- in the pre-classical period; in the Latin of Gaul -ll- passed to -l- after long vowels, hence Fr. étoile < *stēla < stēlla; cf., on the contrary, Ital. stella, Sp. estrella.
51 See e.g. W. von Wartburg, Die Ausgliederung der romanischen Sprachräume 117 ff. (Bern, 1950).
52 The ALF cannot be trusted in this respect. We have reliable data for Hauteville (cf. Martinet, Description phonologique du parler franco-provençal d'Hauteville, RLR 15.28 ff.), Agen, and probably Toulouse (cf. Jean Séguy 20); at Aries (cf. Hélène N. Coustenoble, La phonétique du provençal moderne en terre d'Arles 92 ff.; Hertford, 1945), r- and -rr-have become [r], but -r- is still [r]; the same shift is documented for Hauteville.
53 Strong (phonemically geminated) r is normal in word initial in large sections of southern Italy (see Rohlfs 278 f.), where it should be accounted for as an analogical extension of the geminate resulting from the assimilation of the final consonant of a preceding word. In Western Romance, strong r does not result from an assimilation of a preceding consonant, but from a reinforcement after a consonant which is preserved: -os ro- would yield -or ro-in Southern Italian, but -os Ro- in Western Romance.
54 P. 202.
55 See e.g. Rohlfs, Le gascon (Beiheft 3), ZfrPh. 85.83 ff. Should we also assume that the Catalan analogical extension of the weak reflex of n- is due to Basque influence?
56 See e.g. F. Sommer, Handbuch der lateinischen Laut- und Formenlehre 155 (Heidelberg, 1948).
57 Bourriez 165, 342.
58 Rohlfs, Le gascon 93 f.
59 In general, and for the fate of medial Brythonic -y-, see Lewis–Pedersen 13 ff.
60 Word-medially, however, -sw- > [-v-]; ibid. 18.
61 Ibid. 11 ff.
62 More exactly, all single consonants that were surrounded by fairly open articulations.
63 In phonetic transcription (between square brackets), italic l indicates the weaker allophone.
64 Les mutations consonantiques du germanique 51–70 (Paris, 1948).
65 Some problems of Italic consonantism, Word 6.28–33.
66 It is true that, in such a case, we would have to explain why central and southern Italy does not share this pattern and, instead of weakening the geminates, is found to preserve them and even increase their number.
67 See Antonio Tovar, Estudios sobre las primitivas lenguas hispánicas, map opposite p. 200 (Buenos Aires, 1949).
68 Cf. however the interpretation of prothetic a before morpheme-initial r in the Alemannic dialects of Valais, by N. Otto Heinertz, Eine Lautverschiebungstheorie 52 (Lund–Leipzig, 1925), a book where Celtic lenition is adduced to account for certain features of High German phonology.
69 Les mutations consonantiques 102.
70 Martinet, Description phonologique 40 ff.; Carl Borgstrøm, Zur Phonologie der norwegischen Schriftsprache, NTS 9.256 f.
71 See Vendryes, La place du verbe en celtique, MSL 17.337 ff.
- 4
- Cited by