1In the following I will discuss the Breton word for ‘miller’, meliner, miliner. At first sight we appear to be dealing with a completely regular agent noun in ‑er (from Lat. ‑ārius) derived from melin, milin ‘mill’. However, a thorough investigation of the relevant forms (e.g. MBret. melinhezr, Haut-Vannetais melinaire) reveals that this is not the case. Rather, the word for ‘miller’ is formed, in both Breton and Cornish, with the rare agent noun suffix deriving from Latin ‑ātor.
- 1 The shortening of pretonic ā > ă will not concern us here, as it has no bearing on the problems d (...)
2Before a discussion of the attested forms of ‘miller’, I will first provide some background to the history of the Latin agent noun suffix ‑ātor in the British Celtic languages. Both the Latin nominative ‑ˈātor and the oblique *‑ăˈtōr‑1 (probably accusative *‑ătōrem) are reflected in the British Celtic languages, as *‑ɔdr and *‑adʉr. These two suffixes *‑ɔdr and *‑adʉr were probably still simple allomorphs in a LPBrit. two-case system (subject vs. oblique). With the break-down of the case system, the allomorphy was variously handled in the individual branches:
-
Welsh preserves both variants of the suffix (‑awdr, ‑adur), often interchangeable and with no functional difference (cf. Russell 1989: 40-42).
-
- 2 All of the Old Cornish material in this article is quoted from Graves’ edition of the Vocabulariu (...)
Cornish has generalized the oblique form (‑adur) with one generally recognized exception, OCorn. cusulioder ‘counselor’.2
-
With a single possible exception (OBret. pascatur, DGVB 281-282), Breton has generalized the nominative form (‑ezr), probably in order to avoid homophony with the highly productive action noun suffix ‑adur (< Lat. *‑ătūra or more probably later Gallo-Rom. *‑adʉra). The further development of the Breton reflex is obscured in the Modern Breton period by the regular merger of final ‑ezr with ‑er (from Lat. ‑ārius) in most dialects (see § 3.3 below for an important exception).
3The words formed with this suffix may be conveniently divided into three groups: a) certain or probable direct borrowings from Latin (of whatever kind), b) new British Celtic formations using Latin lexical material, and c) new British Celtic formations using native lexical material.
4The following list presents the more reliable instances of direct borrowings from Latin. However, in many cases, the accompanying verb and other nominal formations were borrowed alongside the agent noun in ‑ātor. So, while *pexɔdr, *pexadʉr ‘sinner’ are almost certainly from Latin peccātor, *peccătōrem, the fact that the verb *pex‑ ‘to sin’ and the action noun *pexɔd ‘a sin’ were also borrowed from Latin (Lat. peccāre, peccātu‑) means that we cannot be absolutely sure of this. Theoretically, most of the forms in group a) could belong to group b). However, in light of the very limited productivity of this suffix in SWBrit., it seems far more probable that we are dealing with direct borrowings of the agent noun from Latin.
- 3 Of course, a direct borrowing of Lat. cō(n)siliārius is also possible for MBret. cusulyer. Howeve (...)
- 4 MBret. empalazr, impalazr ‘emperor’ must be a later borrowing, after the establishment of the new (...)
- 5 It appears that i‑affection has been generalized in this root in Cornish.
Lat. cō(n)siliātor ‘counselor’ > OCorn. cusulioder /ky'syljœder/ (recent formations in MW kyssulywr, MBret. cusulyer3).
Lat. creātor ‘creator’ > MW creawdyr (ML 154; a recent formation in MBret. cro(e)er).
Lat. imperātor ‘emperor’ > MW ymherawdr (ML 132).4
Lat. ōrātor ‘speaker, orator’ > MW arawdyr (ML 133).
Lat. peccātor ‘sinner’ > MW pechadur, MCorn. peghadur, MBret. pechezr (ML 194).
Lat. piscātor ‘fisherman’ > OCorn. piscadur, MBret. pesquezr (more recent formations in MW pyscodwr and ModBret. pesketaer).
Lat. salvātor ‘saviour’ > MCorn. sylwadur5 (ML 209; a recent formation in MBret. saluer).
Lat. senātor ‘senator’ > MW henadur (if not in group c) from hen ‘old’).
Lat. vēnātor ‘hunter’ > MBret. guinhezr (ML 175-176; Bret. ‑i‑ for expected ‑oe‑ unexplained).
5The following list contains instances of the suffix spreading to already borrowed Latin words. The list is intended to be complete for SWBrit.
- 6 This is the traditional etymology advocated by e.g. Loth (ML 155) and Pedersen (VKG I, 198). An a (...)
Lat. disco ‘to learn’ > MW dyscu ‘to learn; to teach’ → dysgawdyr ‘teacher’.
Lat. fuga ‘flight’ > MW ffo ‘flight, escape’, ffo‑ ‘to flee, escape’ → MW ffoawdyr, ffoadur ‘fugitive’.
Lat. pāsco ‘to pasture; to feed’ > MW pasgaf, MBret. pascaff → MW pasgadur ‘herdsman’, OBret. pascatur gl. epulo (cf. DGVB 281-282 for a discussion of this somewhat problematic gloss).
Lat. quiēsco ‘to rest, to keep quiet; to sleep’ (apparently through *kwɪsk‑, cf. OFr. coi ‘calm’ < *kwẹ̄tu‑ < CLat. quiētu‑) > MW cysgu, OCorn. cuske → MW cysgadur ‘sleeper’, OCorn. cuscadur.6
Lat. rēmus ‘oar’ > OCorn. ruif → ruifadur ‘rower’.
6This last category shows a number of instances of spread of the suffix to inherited Celtic material. Interestingly, this category only contains examples from Welsh.
MW cof ‘memory’ → cof(i)awdyr, cof(i)adur ‘one who remembers, secretary, chronicler’.
MW adeil ‘building’, adeil(i)‑ ‘to build’ → adeilawdyr ‘builder’.
MW pen(n) ‘head’ → pennadur ‘ruler, chief’.
MW llyw(i)o ‘to rule, to steer’ → llyw(i)awdyr ‘ruler, governor, pilot’.
7In the following section I present the examples of Breton agent nouns formed with the suffix ‑ezr, the regular reflex of Brit. *‑ɔdr from Lat. ‑ātor (there are no known reflexes of *‑adʉr < *‑ătōrem in Middle and Modern Breton). This will allow us to properly evaluate the significance of the spelling <‑ezr> in the Middle Breton word melinhezr (§ 3.2).
MBret. pechezr : Cms 122v, 122v (fem. ‑es) ; Ca 154, 154 (fem. ‑es) ; M 41 (pl. ‑ien), 159, 451, 515 (pl. ‑yen), 548, 595, 595 (fem. ‑es), 799 (pl. ‑ien), 811, 816, 839, 839 (fem. ‑es), 1135 (pl. ‑ien), 1143, 1171, 1176 (pl. ‑ien), 1323 (pl. ‑ien), 1149 (pl. ‑ien), 1379 (pl. ‑ien), 1471 (pl. ‑ien), 1611 (pl. ‑ien), 1619 (pl. ‑ien), 1638 (pl. ‑ien), 1639 (pl. ‑ien), 1651, 1844 (pl. ‑ien), 1970 (pl. ‑ien), 2203, 2465 (pl. -ien), 2670 (pl. ‑ien), 3583 (pl. ‑ien) ; J 87 (fem. ‑es) ; Kerampuil’s 1576 Catechism 2v, 3v, 7v, 17v, 1**v (pl. ‑yen), 3**, 3** (fem. ‑es), 3**v, 3**v (fem. ‑es) ; H 17 (rhymes with saluer), 45 (rhymes pechezr pechezres) ; Doctrin an christenien (1622) 23 (pl. ‑ien); Saint Yves (1623) 163, 163 (spelled pecherz), 180.
MBret. pecher: H 11 (rhymes with mecher); Nl. 243 (pl. ‑ien), 298 (pl. ‑yen) ; J 11 (pl. ‑yen), 14 (fem. ‑es), 43 (pl. ‑yen), 91 (pl. ‑yen) ; Cant. Spir. (1642) 28 (pl. ‑ien), 33, 60, 79 (pl. ‑ien), 105, 124, 130 (pl. ‑ien), 134 (pl. ‑ien), 147.
MBret. pesquezr : Cms 124v ; Ca 156 ; Nom. 174, 174, 313, 318 ; derived action noun pesquezres (sic, for pesquezrez) Cms 124v ; pesq̄zrez : Ca 156a.
MBret. pesquer* : Nom. [341] (pl. ‑yen).
Place names (Deshayes 1999 : 195) : Querpesquezre (1427 ; Kerbisquer, Sarzeau) ; Keranpesquezrien (1540 ; Kerbesquerrien, Cléden-Cap-Sizun).
MBret. guinhezr : Ca 106a ; Cb fol. L ; guīhezr Cms 95v ; gumherz (r. guinherz) Quiquer’s Dict. et Coll. (1626, glossing chasseur in the dictionary section) ; guinezr Nom. 317 ; guiznezl Nonne before line 273/396 ; guinhezl Nonne 273/396 (rhymes guinhezl quehezl).
Family names (Deshayes 2005 : 248) : Guinhezr (1497, Morlaix), An Guynerre (id.), Le Guyner (1499, Ploaré), Le Guynhezre (1531, Morlaix), Le Guinezre (1604, Ploujean).
8As can be seen, the spellings with <‑ezr> predominate, especially in the earlier attestations. In most cases, the spellings with <‑er> probably represent a genuine reduction of the cluster ‑zr to ‑r. However, we should not be blind to the possibility of suffix substitution, considering the popularity of the isofunctional suffix ‑er (< Lat. ‑ārius). Another factor to be taken into account, especially with regard to family and place names in non-Breton sources, is the possibility of “French” spellings. Since the French orthography at the time had no way of writing /δ/, attestations such as An Guynerre, Le Guyner and, indeed, Le Melinnaeyre (§ 3.2 below) might simply be French renditions of the Breton words with /‑eδr/. With this in mind, we may proceed to the attestations of the word for ‘miller’ in Breton.
9ModBret. miliner, meliner has until now been derived from earlier *mölinœr and ultimately considered a borrowing from Latin molīnārius. However, there is nothing in the Modern LTK Breton attestations to tell us whether the suffix ‑er derives from MBret. ‑er or ‑ezr. These attestations are therefore of no value in determining whether we should posit earlier ‑ezr (Lat. ‑ātor) or ‑er (Lat. ‑ārius) for this word, and we have to turn to the older stages of the language to achieve clarity.
10The word for ‘miller’ is attested in Old Breton as molinor gl. pistor ‘miller’ /mölinœr/ (DGVB 259). The knee-jerk reaction would be to derive this directly from Latin molīnārius ‘miller’ and to try to make the remaining attestations from Breton (and Cornish) fit with this assumption. However, in light of the evidence presented below, this is not possible. As will be argued, we may have to interpret molinor as a recent, synchronically regular derivative of OBret. molin /mölin/ ‘mill’, with the productive suffix ‑or /‑œr/.
11The word for ‘miller’ is poorly attested in the Middle Breton literature:
MBret. melinhezr : Ca 134b.
MBret. miliner : Nom. 314.
Family names (Deshayes 2005 : 264) : An Melinhezre (1426, Carnoët), Le Melinnaeyre (1447, Guidel), Le Melinezre (1514, Saint-Avé), Melinhezre (1541, Plogoff), Miliner (1576, Trégarantec), Le Meliner (1630, Melgven).
- 7 “ezr for er: bouclezr buckler, melinhezr miller, salinhezr salt-cellar.” (Stokes 1876 : 76).
- 8 “Im Nbr. mußte dies Suffix [MBret. ‑ezr < Lat. ‑ātor] mit lat. ‑ārius [> MModBret. ‑er] zusammenf (...)
- 9 “molinor (inédit, BN lat. 10290, fo 38a; […]) gl. « pistor », « cuisinier, boulanger ». Bret. moy (...)
12The MBret. forms in ‑ezr(e) are usually interpreted as having unetymological, presumably hypercorrect ‑zr. This was the opinion of Stokes,7 Pedersen8 and Fleuriot.9 However, the persistent ‑ezr both in the Catholicon and in family names from various sources (and places) stands out like a sore thumb. The word acts just like the other words with etymological ‑ezr, such as pesquezr, pechezr, guinhezr (§ 2 above). Can it really be hypercorrect, as has thus far been assumed?
13Unetymological, presumably hypercorrect, <zr> is extremely rare in Classical Middle Breton (e.g. Cms, Ca, B, M, N, J, Pm. and earlier documents). This will be shown by a complete list of instances, with a brief etymological commentary:
14<zr> etymologically justified:
arazr ‘plough’ (W aradr, OIr. arathar), azr ‘snake’ (W neidr, OIr. nathir), azrec ‘compunction’ (W atreg, OIr. aithrech), azrouant, pl. ezreuent ‘enemy, demon, devil’ (probably comparable to MW edrywant ‘misery’), bezret ‘cemetery’ (W beddrod), breuzr ‘brother’ (OIr. bráthair), calizr ‘chalice’ (< *kalɪgr < Lat. caliculus; OCorn. kelegel), cazr ‘beautiful’ (W cadr), cazrell ‘weasel’ (← cazr ‘beautiful’, cf. Fr. belette ‘weasel’), coazrell ‘sole (of a shoe)’ (< EOFr. *kwaδrεll (OFr. quarrel) < VLat. *quadrellus), commazr ‘fellow sponsor (fem.)’ (< VLat. *commater), compazr (misspelled comparz in Ca), pl. compizrien ‘godfather, fellow sponsor (masc.)’ (< VLat. *compater), croezr ‘sieve’ (W crwydr), dazrou ‘tears’ (W dagrau), dezreuell and ezreuell ‘to tell’ (cf. OBret. etlebellet), dezrou ‘beginning; to begin’ (W dechrau), empalazr, impalazr ‘emperor’ (VLat. *ẹmperator), enquelezr ‘giant’ (OCorn. enchinethel), gozro ‘to milk’ (MW godro), feuzr ‘sheath’ (Germ. *fōdr), goezreu “cest vne maladie des yeulx” (Ca; *goez ‘sight’ + reu ‘frost’?), gouzroncquet ‘to bathe’ (< *u̯o‑tronk‑etu‑, cf. MCorn. tronkys, OIr. fothrucud), guezr ‘green’ (< *gwerδ < VLat. vir’dis), guezrenn ‘a glass’ (< Lat. vitrum), guinhezr ‘huntsman’ (< Lat. venātor), hezr ‘bold’ (W hydr), hezreff ‘October’ (W hydref), lazr, pl. lazron ‘thief’ (W lleidr < Lat. latrō), leuzr ‘pole (of a vehicle)’, leuzr‑ ‘to lead, conduct’ (OIr. láthar ‘position’), lezr ‘leather’ (W lledr), louazr ‘trough’ (OIr. loathar), louzr ‘stocking’ (W llawdr), mazron ‘godmother’ (< VLat. *matrọna < CLat. mātrōna), milhezr ‘maladie des pieds’ (MW maleithr), mozreb ‘aunt’ (W modryb), pazron ‘godfather’ (< VLat. *patrọn‑ < CLat. patrōnus), pechezr ‘sinner’ (< Lat. peccātor), pesquezr ‘fisherman’ (< Lat. piscātor), Pezr (< Lat. Petru‑) alongside Piezres (Kerampuil’s 1576 catechism (12*); Cart. Landev. Alanus Piezresii (1); < EOFr. *piεδrəs < Lat. Petrus) and Pezron ‘Peter’ (< EOFr. *pεδrọn, cf. OFr. obl. Perron), scuezr ‘square, frame’ (< EOFr. *(e)skwẹδr < VLat. *exquadra), squezrenn ‘splinter’ (probably comparable to W ysgythr‑ ‘to cut or hack off’), tarazr ‘auger, drill’ (W taradr).
15Etymology unknown:
clezrenn ‘ice’, clo(e)zrenn ‘pod, husk’, sclezrenn ‘scraping tool’ (possibly instrument nouns in ‑zr).
16<zr> not etymologically justified (see § 6 for a discussion of these words):
trichezr ‘trickster’ (OFr. tricheor), lichezr ‘lecher’ (OFr. licheor), salinhezr ‘container for salt’ (MFr. saliniere), bouclezr ‘shield’ (OFr. boucler).
17A few words with ‑zr are only attested from late Middle Breton or early Modern Breton texts onwards, yet probably continue a real, orthographical tradition, e.g. feuzr ‘fur, stuffing’ (Nom.; OFr. fuerre < Germ. *fōdr‑), clezr ‘bars, rails’ (Grég.; cf. W cledr). However, the value of these examples is greatly diminished by the fact that the same texts also show reduction of ‑zr > ‑r and, as a consequence of this, hypercorrect ‑zr‑ (e.g. quantolezr for quantoler already in the Nomenclator).
18Vannetais Breton, with its predominant (if no longer universal) final stress, preserves distinctions in the final syllable that were lost in LTK Breton before or in the course of the Middle Breton period. In the case of the word for ‘miller’, it turns out that Vannetais provides remarkable information supporting the authenticity of MBret. ‑zr.
19First I will give a list of attestations of ‘miller’ in early Haut-Vannetais texts:
Châlons Dict. Bret.-Fr. (1723): meliner, fem. melineres (119; inconclusive)
L’A*** Dict. Fr.-Bret. (1744): melinaire, pl. melinairion (239)
Magasin spirituel er beurerion (1790): melinnær (V, 81, 82, 124, 211, 212, 213, 220), pl. er velinnærion (211, 212)
Vocabulaire nouveau, ou dialogues français & bretons (first edition, 1790s) : ur Melinær (38), fem. ur Velinærès (38).
Addition au Vocabulaire Français-Breton (1800-1810 ?) : melinêr, melinêrïon (25).
20As can be observed, HVann. ‘miller’ is spelled with <‑aire>, <‑ær>, <‑êr>. These spellings represent an open vowel /‑εr/. This may be compared to the following list of attestations of agent nouns corresponding to Middle Breton ones in ‑er (< *‑œr < Lat. ‑ārius ; MBret. glaoüaer, guiader, labourer, porz(y)er, quemener, saluer, toer) :
Châl. : gùiadér (94), guiadér (155) ; labourér (107) ; pobér (13, 135) ; porhér (137) ; queminér, quemenér (140).
L’A*** : gléuaére (53) ; gùiadér (383) ; labourérr (212) ; pobér (35) ; porhérr (295) ; queminérr (375) ; salvérr (351) ; touér (83).
Mag. Spir. : labourér (326) ; queminér (v, 247), ur heminér (340, 341, 395) ; salvér (102, 255, 270, 283, 437).
Voc. Nouv. : gleuaér (44) ; labourér (7) ; pobér (37) ; porhér (44) ; queminér (183), ur heménér (41) ; salvér ([1]) ; touér (43).
21As can be seen, in these instances the suffix is consistently spelled <‑ér>, <‑érr>, <‑ére>, i.e. it has a close vowel /‑er/. One way to account for the unique /‑εr/ in ‘miller’ would be to take MBret. melinhezr seriously. Early MBret. *melinœδr, *melineδr (vel sim.) would account for the open vowel of HVann. melinaire (cf. *brœδr > braire ‘brother’, *leδr > laire ‘leather’). I am unaware of any alternative explanation for the peculiar behaviour of the suffix in melinaire.
22The evidence from early Haut-Vannetais is mirrored in the modern Vannetais dialects, as shown by the following collection of data from the ALBB.
|
ALBB
|
60
|
61
|
62
|
63
|
64
|
65
|
67
|
68
|
69
|
70
|
|
458 ‘miller’
|
[‑ˈεːr]
|
[‑ˈiːər]
|
[‑ˈεːr]
|
[‑ˈεjr]
|
[‑ẹːr]
|
[‑ˈẹːr]
|
[‑εːr]
|
[‑ˈεːr]
|
[‑ˈẹːr]
|
[‑ˈẹːr]
|
|
299 ‘weaver’
|
[ˈ‑ər]
|
[ˈ‑ẹr]
|
[ˈ‑er]
|
[‑əjr]
|
[‑ˈεr]
|
[‑ˈẹjr]
|
[‑ˈẹːr]
|
[‑ˈiːr]
|
[‑ˈiːr]
|
[‑ˈεːr]
|
|
372 ‘taylor’
|
[ˈ‑εr]
|
[ˈ‑ẹ]
|
[‑ˈẹːr]
|
[‑ˈεjr]
|
[‑ˈer]
|
[‑ˈεr]
|
[‑ˈεːr]
|
[‑ˈiːr]
|
[‑ˈiːr]
|
[‑ˈẹːr]
|
|
71
|
72
|
73
|
74
|
75
|
76
|
77
|
78
|
79
|
81
|
82
|
83
|
|
[‑ˈεːr]
|
[‑ˈεːr]
|
[‑ˈεr]
|
[‑ˈεr]
|
[‑ˈeːr]
|
[‑ˈεr]
|
[‑ˈεːr]
|
[‑œˈεːr]
|
[‑εːr]
|
[‑εːr]
|
[‑ˈεr]
|
[‑ˈεːr]
|
|
[‑ˈεr]
|
[‑ˈẹːr]
|
[‑ˈiːr]
|
[‑ˈįːr]
|
[‑ˈiːr]
|
-
|
[‑ˈiːr]
|
[‑ˈiːr]
|
[‑ˈiːr]
|
-
|
[‑ˈεːr]
|
[‑ˈεːr]
|
|
[‑ˈεːr]
|
[‑ˈεːr]
|
[‑ˈiːr]
|
[‑ˈir]
|
[‑ˈiːr]
|
-
|
[‑ˈiːr]
|
[‑ˈiːẹr]
|
[‑ˈiːr]
|
[‑ˈiːr]
|
[‑ˈεːr]
|
[‑ˈẹːr]
|
23Note: map 299 ‘weaver; tisserand’ represents both gwiader and teiser, map 372 ‘taylor; tailleur’ represents both kemener and gwrier.
24The material is unclear for a number of points (especially when the contrast *εː vs. *eː hasn’t been shifted to /εː/ (/eː/) vs. /iː/), but it is in general clear for southern Haut-Vannetais, pts. 68, 69, 73, 74, 75, 77, 78, 79, 80, where we find a contrast between [‑ˈεːr] ([‑ˈeːr]) in meliner and [‑ˈiːr] in the other agent nouns (from older *‑εr vs. *‑er).
25In the Bas-Vannetais dialect of Île-de-Groix, the word for ‘miller’ has the form /məlinaer/ (Ternes 1970: 159, 195, 333), whereas other agent nouns corresponding to MModBret. ‑er end in /‑eir/ (332-333). Thus we find the same situation as in Haut-Vannetais, since /‑aer/ points to *‑εr (< *‑œ(δ)r, *‑ai(δ)r), while /‑eir/ points to *‑er. Note that *‑eδr probably won’t work as a protoform, since this gives /‑eaːr/ (/peaːr/ ‘Peter’ (10) < *peδr, /leaːr/ ‘leather’ (11) < *leδr, /dʒwaːret/, lenited /ərɥeaːret/ ‘cemetery’ (129, 157) < *gweaːred < *beaːred < *beδrœd) or /‑eir/ (following a palatalized stop).
26We now have several indications of the authenticity of MBret. ‑zr in melinhezr: the occurrence of <zr> in both the Catholicon and Middle Breton family names, and the open vowel reflex in early Haut-Vannetais literature, modern Haut-Vannetais dialects and Île-de-Groix Breton (a vowel reflex which is distinctly lower than the one found in other agent nouns in ‑er). But we still have no smoking gun. The Cornish material will provide this.
27The word for ‘miller’ is not attested in Cornish literature as such, but we have a good idea of its shape from other sources:
Middle Cornish pers. epithets (quoted from Padel 1985: 161): Thomas Melyneder (1327), William Melender (1562).
Late Cornish: belender (AB 14a, 93a, 240c).
Place names: Park Belender, Crauft an Mellender (field names; Padel 1985: 161).
- 10 “Accretion of [d] after [n] occurred in the word for ‘miller’, originally *melyner; e.g. the surn (...)
- 11 “*melynder (m.) ‘miller’ […] a derivative of melin1, cf. Welsh melinydd […] and Breton miliner: t (...)
- 12 In his discussion of the phenomenon of mistaken “radicalization” of lenited v‑ > b‑ for etymologi (...)
28Cornish ‑nd‑ instead of ‑n‑ (as would be expected if from *mölinœr) has thus far remained a mystery, cf. the discussions by George,10 Padel11 and Chaudhri.12 However, an important fact appears to have gone unnoticed. Between the oldest attestation, Melyneder (1327) and the second-oldest, Melender (1562), something has happened: syncope of the penultimate vowel. This only happens in Cornish if the penultimate vowel is irregularly unstressed, and the only way that the penultimate vowel can be unstressed in Cornish (excluding prepositional compounds etc.) is if it – or the following vowel – has developed by epenthesis only after the fixing of the accent on the penultima. We shall therefore take a closer look at epenthesis and syncope in Cornish.
29Vocalic epenthesis is regular in various clusters of consonant (stop, δ) + sonant (r, l, n). In a previously monosyllabic words (exemplified here by words ending in *‑dr), the epenthesis does not violate any rules of stress placement:
‑ˈVdr > ‑ˈVder
- 13 Likewise in Old Cornish (examples from the Vocabularium Cornicum): cader (ms. carder; MBret. cazr (...)
SWBrit. *ˈnadr > OCorn. ˈnader > MCorn. ˈnader (2 sylls.; MBret. azr)
SWBrit. *ˈladr > OCorn. ˈlader > MCorn. ˈlader (2 sylls.; MBret. lazr)
SWBrit. *ˈbrɔdr > OCorn. ˈbroder > MCorn. ˈbroder, ˈbruder /ˈbrøder/ (2 sylls.; MBret. breuzr)13
30As can be seen, even with the appearance of a new syllable, the placement of stress is still regular.
31In words which, prior to epenthesis, had two or more syllables, the epenthetic vowel created a problem. The stress was retracted to the penultima before the epenthetic vowel arose (or at least, before it was phonemicized):
‑VˈCVdr > ‑ˈVCVdr (penultimate stress) > ‑ˈVCVder (epenthesis)
32This created a situation where the stress was on the ante-penultimate syllable. Regular stress on the penultima was restored by syncopating the middle vowel (cf. the dialectal Breton ˈkendro < *ˈkendero < *ˈkenderu̯):
‑ˈVCVder > ‑ˈVC’der (syncope)
33This syncope can be observed in the following examples:
PCelt. *aratro‑ > *ˈaradr (MBret. arazr, MW aradyr) > *ˈarader (VC aradar, arader-uur ‘plough-man’) > LCorn. ardar ‘plough’ (AB 7b, 43b, 155a, 241a, 290a).
PCelt. *teratro‑ > *ˈtaradr (MBret. tarazr, OW tarater, MW taradyr) >*ˈtarader > MCorn. tardar* ‘auger, drill’ (ov thardar in OM 1002) > LCorn. tardar ‘auger’ (AB 7b).
PCelt. *u̯elatro‑ ‘prince, chieftain’ >*‑ˈu̯aladr (OBret. ‑uualatr in cpd. names, MW gwaladyr) > *‑ˈu̯alader > Place name Trewaleder (1286) > Trewalder (Weatherhill 2005: 90; cpd. with *treβ‑).
Lat. fenestra ‘window’ > *ˈfenestr (MBret. fenestr, frenest, penestr) > *ˈfenester (VC fenester) > *fenster > LCorn. beisder (AB 12c, 21c), beisdar (AB 59a), dre an veistir (AB 249b).
The personal name *kɵn‑hɪdr > *‑ˈgen(h)ɪdr > Place name Tregeneder (1284) > Tregender (Weatherhill 2005: 109; cpd. with *treβ‑).
- 14 As suggested by the lack of pre-occlusion in Late Cornish.
34Final *‑VCVδl probably developed in the same way. In the case of *‑VnVδl, the secondary cluster *‑n’δ‑ arising by syncope apparently gave MCorn. simple ‑n‑14 (alternatively, one could operate with loss of *δ between post-tonic vowels):
PCelt. *bVnatlo‑ > *ˈbanaδl > OCorn. *ˈbanaδel (VC banathel) > *ˈbanδel > LCorn. banal ‘broom flowers’ (AB 63a, 272c), bannolan (AB240c), dial. bannal (Padel 1985: 16).
PCelt. *mVnatlo‑ > *ˈmanaδl > OCorn. *ˈmanaδel > *ˈmanδel > LCorn. manal ‘sheaf’ (AB 33a, 241a).
PCelt. *anatlo‑ > *ˈanaδl > OCorn. *ˈanaδel > *ˈanδel > MCorn. *ˈanel → vb. anella ‘to breathe’ (TH 2, 36, 38a).
35And finally, an example with a fairly rare cluster, probably showing the same development:
- 15 Cf. the epenthesis in OCorn. kelegel gl. calix < *kẹlɪgl < VLat. *calic’lu‑ < Lat. caliculu‑. The (...)
LPBrit. *ˈhʉδɪgl ‘soot’ (MW huδygl, MBret. huzel) > OCorn. *ˈhyδɪgel15 > *ˈhyδgel > *hylgeδ (metathesis) > LCorn. filgeth (AB 33a; with f‑ for expected h‑ apparently due to an exceptional phonological development). This path seems more straightforward than the twice irregularly metathesized *hyδegl > *hygleδ > *hylgeδ suggested by Wmffre (2003: 331).
36As can be seen from these examples, epenthesis followed by syncope is a completely regular development. It is also the only way to account for the syncope in the Cornish word for ‘miller’. The development must have been as follows:
OSWBrit. *möˈlinVdr > Meˈlyneder (1327) > Meˈlender (1562) > LCorn. beˈlender
37We can now summarize the findings of the preceding sections:
-
Middle Breton has ‑ezr and not just ‑er as we see in agent nouns from *‑œr.
-
Early Haut-Vannetais has ‑aire, ‑êr, ‑ær and not **‑ér as we see in agent nouns from *‑œr.
-
Modern Vannetais dialects usually have a distinct open vowel reflex in the suffix of the word for ‘miller’ as opposed to the suffix of other agent nouns corresponding to MBret. ‑er.
-
The dialect of Île-de-Groix has /‑aer/ and not **/‑eir/ as we see in agent nouns from *‑œr. Additionally, the final /‑aer/ in this dialect may exclude older *‑eδr and point specifically to *‑œδr.
-
MCorn. Melyneder, Melender, etc. can only be explained from older *mölinVdr.
38Thus we can conclude that the attested forms (with the notable exception of OBret. molinor) are incompatible with the traditional reconstruction *mölinœr. Instead they seem to point to SWBrit. *mölinœdr, as if from Lat. *molīnātor.
39In order to try to understand how this *mölinœdr came to be, we shall now take a second look at the probable instances of productivity of the British Celtic suffix *‑ɔdr, *‑adʉr in SWBrit.:
OCorn. cuscadur ‘sleeper’ ← *kusk‑ (OCorn. cuske ‘to sleep’) < Lat. quiēscere ‘to rest; to sleep’.
OBret. pascatur gl. epulo ← *pask‑ (MBret. pascaff) < Lat. pāscere ‘to pasture; to feed’.
OCorn. ruifadur ‘rower’ ← *ruiμ (OCorn. ruif, MBret. roeuff) < Lat. rēmus ‘oar’.
To which we can add:
MBret. melinhezr, MCorn. Melyneder < *mölinœdr ← *mölin < Lat. molīnum.
40None of these formations (cuscadur, pascatur, ruifadur, *mölinœdr) conforms to the existing Latin rules of derivation, which usually require agent nouns in ‑ā‑tor to be derived from verbal stems in ‑ā‑. Both OBret. pascatur and OCorn. cuscadur are derived from verbs of the wrong conjugation (pāscere, quiēscere) and *mölinœdr and OCorn. ruifadur are derived from nouns (molīnum, rēmus).
41The immediate solution would be to take these words as rare instances of productivity of the suffix *‑ɔdr, *‑adʉr in British Celtic (cf. its later productivity in Welsh), and this may well ultimately be correct. However, it does seem like a strange coincidence that the four known instances where the suffix *‑ɔdr, *‑adʉr has spread in SWBrit. are formed to Latin words already borrowed into SWBrit. Therefore one could consider the possibility that these formations represent mistaken attempts by speakers of British Celtic at sounding Latin: using Latin roots and suffixes, but ignoring the proper rules of derivation.
42Alternatively, we could be witnessing a dialectal and otherwise unrecorded spread of the suffix ‑ātor in the variety of Latin responsible for these loan words in British Celtic. The suffix already enjoyed some productivity beyond its original domain in Classical Latin, e.g. salīnātor ‘dealer in salt’ (← salīnae ‘salt-works’), gladiātor ‘swordsman, fighter’ (← gladius ‘sword’), where no verbs in ‑ā‑ were present. It might therefore merely be the case that (a variety of) spoken British Latin produced formations such as *quiēscātor, *pāscātor, *rēmātor and *molīnātor.
43If a form *mölinœdr ‘miller’ must be reconstructed for SWBrit., as argued above, where does that leave OBret. molinor? As I see it, we have at least two choices:
-
OBret. molinor /mölin‑œr/ may be a perfectly regular recent formation based on molin /mölin/ with the highly productive suffix ‑or /‑œr/ (< Lat. ‑ārius). Cf. MW melinyδ (with the productive suffix ‑yδ), Bret. dial. milinour, melinour (with the productive suffix ‑our < *‑u̯ur and WFr. ‑(ë)our). This would make the similarity of OBret. molinor and Lat. molīnārius coincidental. Cf. also the likely replacement of *‑œdr/‑ezr with productive *‑œr/‑er in Lat. creātor (MW creawdr) → MBret. croeer (not **croeezr) and Lat. salvātor (MCorn. sylwadur) → MBret. saluer (not **saluezr).
-
OBret. molinor really is from Lat. molīnārius, and the reflexes of molīnārius and *molīnātor existed side by side dialectally in SWBrit. and later in OBret. Theoretically, molīnārius may underlie the Modern LTK forms meliner, miliner as well, but not MBret. melinhezr, nor the Vannetais forms. Note that it would be difficult to make this hypothetical dialectal division fit the map, seeing as a substantial part of the Breton linguistic area is already covered by the reflex of *molīnātor.
44As was noted above (§ 3.2.1), there are a few instances where we appear to have hyper-correct <zr> in classical Middle Breton. These will be discussed below.
MBret. trichezr ‘trickster’ (Ca 202a, Cc, M 595, 839), tricher (Cb), abstract derivative trichezry (Cc, M 811, 1143, 1171), trichezri (M 1651), trichery (Cb), adjective trichezrus (Cb). Rhymes with words in <ezr> in M.
MBret. lichezr ‘lecher’ (Ca 24a, 125b, Cb 93a, Nom. 324b), licher (Cb 93a), feminine lichezres (Cb, DÉBM 236), abstract derivative lichezri (Ca 125b), lichezry (Cb 93a), lichery (Cant. Spir.: 107), lichizry (Cb 92a), derived verb lichezraff (Cb 58b).
Personal name: Le Lychezr (1540; GMB 367).
45According to Piette (1973: 186), in trichezr and its derivatives “the ‑z‑ is of course a mere imitation of B[reton] spelling, cp. bouclezr, etc.”. Similarly, Piette (1973: 139) remarks that in lichezr, the “‑z‑ is of course a mere imitation of MB. spelling, as in bouclezr”. What is meant by this is apparently that <zr> was merely an orthographic archaism in classical Middle Breton, used only due to tradition. However, as the evidence clearly shows, there can be no doubt about the authenticity of <zr>. It seems inconceivable that these two words and their derivatives should be spelled with <ezr> almost without exception in different sources and rhyme with words in /eδr/ if they didn’t actually contain /eδr/. Therefore we may take a second look at the etymology.
- 16 MBret. Piezres, scuezr given in § 3.2.1 probably belong to the same stratum of loan words.
46While the words are obviously borrowed from some variety of French, a hint about their exact origin is given by the French translation in the Catholicon, where we find trichezr translated by “tricheur ou tricherre mauuais” and lichezr translated by “licherre”. The Catholicon French forms tricherre and licherre (instead of the expected MFr. tricheur, licheur) in all likelihood reflect a dialectal continuation of the stem of the nominative singular of the Old French case system (OFr. trichierre, lichierre). The nom. sg. suffix ‑erre derives from early Old French ‑edre /‑eδrə/ (incidentally also from Lat. ‑ātor). While the nom. sg. in ‑edre happens not to be attested for tricherre and licherre, it must have existed, and could very well have been borrowed by Breton, thus explaining the MBret. forms with ‑ezr.16
- 17 E.g. nom. sg. prestres, acc. sg. prestre from older nom. sg. prestre, acc. sg. provoire (Spence 1 (...)
47However, since it is unlikely that the Old French case system had collapsed before the reduction of /δr/ in early Old French, we cannot explain MBret. trichezr, lichezr as borrowed from *trichedre, *lichedre (uninflected for case). Therefore, we must rather be dealing with the generalization of the nom. sg. stem alternant, a process described by Spence (1965: 231-243). This process, observable for other irregular stem classes,17 would produce a paradigm nom. sg. *trichiedres (with nom. sg. ‑s), acc. sg. *trichiedre, which in turn can explain both the Middle Breton forms in ‑ezr (as early borrowings) and the Catholicon French forms in ‑erre. Needless to say, other explanations may eventually present themselves.
MBret. salinhezr ‘container of salt’ (Ca 179), saliner (Coll. Fr.-Br.), salliner (Nom. 157a), EModBret. saligner (Maun.). From Old or Middle French saliniere (a recent derivative of saline or a borrowing from OProv. salinier).
MBret. bouclezr ‘shield’ (Ca 23a, Cb fol. XI), bouclær (Nom. [183]). From OFr. boucler.
48In these two words, it is clear that the MBret. <zr> is not etymologically justified. However, it is important to distinguish between words that are only attested once with <zr> and words which are attested several times with <zr> in different sources. Such a distinction will allow us to weed out accidental misspellings by scribes and printers. Furthermore, a word that occurs at the same place in the various editions of the Catholicon and is spelled identically should naturally only be treated as one attestation, since later editors may simply have copied the older spelling. In the case of salinhezr and bouclezr, the spelling with <zr> is confined to the Catholicon. The lack of sufficiently old attestations outside this work means that we can’t exclude that <zr> is simply a mistake. Therefore, I find it justified to disregard salinhezr and bouclezr as evidence for hypercorrect <zr>.