104

II.-TEE AND THE OTHER ARYANS OF THE P AND Q GROUPS. By JOHN RHYS.

[Read Febrtwry 20, 1891.1

I. IT is a commonplace of Celtic phiIoIogy that the CeItic languages of modern times divide themselves into two groups, namely Goidelic and Brythonic, and that the Goidelic group consists of the Gaelic dialects of Ireland, Man, and Scotland, while the Brythonic group is made up of Welsh, Old Cornish and Breton. It is equally well known that Brythonic speech is characterized by the use of the consonant p (liable accord- ing to its position to become b and ph), while inscriptions in early Goidelic show qv, which is simplified in the later stages of Qoidelic pronunciation into c (liable to be modified accord- ing to ita position into ch and 9). Now the question suggests itself whether the Continental Celts of antiquity had not likewise two languages similarly characterized by p and qti respectively: in other words, whether, besides the , whose language is known to have had p for original p, there was not on the Continent a Celtic people whose language was like early Goidelic and resembled it particularly in retaining q?. I have been gradually convinced that there was such ti people, and I should call it Celts as distinguished from Gauls, but that the modern meaning attached to tho word makes this. inconvenient. One has therefore to fall back on the name given by Pliny to their portion of Gaulish territory, namely Celtic@,’and call them from it Celticans, just ae we speak of the inhabitants of Africa and America as Africans and Americans. Celtica was the country of the peopIe who according to Czesar called themselves Cedtce in their own language. He regarded (exclusive of the

1 Nut. IT&. iv. 105 : Gallia omnis comata uno nomine a pellata in tn’a gpu!orurn genera dividitur, amnibus rnaxume distincta; tl Bcalx ad Sequanam elpica, ab eo ad Ganinnam Celticn eademque Lugdunensis, inde d l’yrenaci montis excursum Aquitanica, Aremorica antea dicta.” CELTS AND OTHER ARYASS.-J. RHYS. 105

Province) BB consisting of three distinct regions, one of which situated beyond the was inhabited by the Aquitani, who were wholly or mainly non-Celtic. The other two peoples were the Celte and the Belp. As to the Celh he 88~8,that they were separated from the Aquitani by the Garonne, and from the Belp by the Seine and its tributary the Marne. That is to say, one is given to understand that Czsar's Celt@, whom it is here proposed to call Celticans, occiipied all North-western, Central and Southern Gaul outside the Roman province ; and within tbe limits of that province itself they probably formed the bulk of the Aryan population there, at least before the Allobroges were annexed to it.' Moreover they had also penetrated into ; for we possess important evidence to their presence there in the well-known name of the mixed people of the Celtiberi. Lastly, as the Celtican element found its way to the heart of the , it may have also possibly reached the North of Italy; but this must depend for ita proof on the interpretation to be put on the slender facts of language to be presently brought forward. Let us begin with the instances of q names found in the ancient inscriptions of the countries here in question, and, for the sake of convenience, let us take first those of Spain and Portugal. Here we have Alhcqtiius, Arquius, Doquiriis, Equaesus and Quarqtcemi, together with some others about which there is less certainty.2 Arquizrs, Alhqrtiu.r. These two names occur in one and the same inecriptions at Valenpa on the Minho in North Portugal, while Alluquitcs occurs elsewhere in an inscription at Pairnogo' in the west of tho old province of BEtica, near the river Guadiana; and somewhat leae certain Is its

1 I make this qualiflcation a8 the name Allobrogrs would seem to mean a people (6 of other marchen :" possibly they belonged to nnother branch of the family. 2 Such a9 ihe dative Gnro Lnquinieti on a stone from Caldas de Vizella near Guimnrrens in the North of Portugnl : see volume ii. of the Bedin Cmpua Itlacrip. Lot. No. 2406. Such also 8s Aflodw Jfaquiaewd Sunnae. F., of doubtful reading : see No. 4980. No. 2465: Dis . Manibus I Alluquio . Andergi . F. I Aetnrae . Arqui . F. I Macro. Alluqui . F. C1 1 utimoni. Alluqui . Y. Civi I Ena...... 4 No. 961 : Glaucua . Aluquii . F. I H.Y.E 106 p AISD Q GROUPS. presence in an inscription at Arroya del Puerco’ not very far from Caceres or the ancient town of Norba in Lusitania. The origin of Alluqicius is obscure; but I cannot help regarding it as possibly one of the forms which Apqtrins took in Iberian This latter name, besides occurring with Alluquius in the Valenpa inscription, is to be met with in others, at Trujillo, the ancient Turgallium, in the east of Lusitania, at Monteagudo between Tarazona and Cascante in the ancient Tarraconensis, and at Astorga, the ancient Asturica Augusta, to the south-weet of Also at Calderuela near Soria between Tarazona and Osme,‘ and at Carripa in the vicinity of Oporto,s besides Arquia in an inacription from Condeixa a Nova, near Coimbra.6 The name Arquiue probably meant ono who had to do with the bow, that is to say, an archer, aid waa derived from a Celtic word7 cognate with the arquuv and arcus: a bow or arch. Whether the adjective Erquesis should be considered as in any way related to Arquiics is extremely doubtful. Doqzcirus, Uocquiricus. We have Doquirus from Trujillo, and Docquirrts from Tdanha a Velha, the ancient Igaeditaiii in central Luaitania, and a probable Doquira12 from Soure in the neighbourhood of Alfeizariio in Lusitania, aleo a trace

1 So. 131 : Ammu. AUuoi. F.,where the reading Aliuyui has been suuggesbd by the editor Dr. Hiihner. * Here my colleaguw Prof. Ncttleship rcminde me of Cieero’s words when in hie oration Pro AreAia he speaks concerning “natia Cordubae poctis, piugue quiddam sonantibus.” Nos. 632, 5990, 2633 respectively. 4 No. 2834. 6 No. 2373 : other iiistances will be found in Nos. 2468, 2438, 243.5, aU from localities in the neighbourhood of Braga, the ancient Bracara Auysk. 6 No. 317. 7 Wo have probably D trace of the word in the Welsh arfed, ‘the lap or the abdomen,’ since arq~must become u~phor mf in that lnnguage : as to the meaning compare the German schambuq. 011 Teutonic ground the word is implied by the Gothic orAcanra ‘ tin arrow,’ A.-Salon sarA, the same. 0 The modern Weluh arch ie tho Latin arcu borrowed, but in the colloquial the English arch (with palatal cA) is usually substituted for it. 9 It comes from Aloala del Rio north of Sevillo, and, purporting to be the atme of a centurion, it occurs in company with such other names as Ucresia, Arvabbresir, Isheaid, Iaurgutuna, ctc. : 6ce No. 1064. lo No. 624. Xo. 448. i2No. 3G4. CELTS AKD OTHER ARYAh’S.--J. RHYS. 107 of the name in an inscription at Alfeizardo itself.’ One may add Docquiricus or Docquiri)ius,z from Freixo de Nemiio on the Douro in Lusitania, Docquirims from Mkrida, the ancient Augusta Emerita in the south-east of Lusitanis, and a doubtful instance4 from Lara de 10s Infantee, south-east of Burgos in Tarraconensis. The origin of Doquirus5 and its derivatives is not certain, but it cannot be Oaulieh ; 80 it may be regarded as here in point, since it has the appesr- ance of being an Aryan word. Equabona. This ie the name of a place in the Itinerary of Antoninus,G and it was on tho route between Lisbon and Mkrida on the Guadiana. It is unmbtakably Celtic, and recalls such other Celtic names as Vindobona and Bononirt, but in Gaulish it would have doubtless been Epobona, like Eporedia and the like, the first element in the compound being forms of the Celtic word which is in Latin eqzius ‘a horse,’ in old Irish ech ‘horse,’ and in old Welsh ep-aul, now ebol, ‘a colt.’ The derivation of bona in Equalona, Vidobonn, and the like, is unknown; it may be of the same origin possibly as the Welsh b6tr ‘ the stem or. trunk (mostly of a tree),’ Irish bun “ Wurzel- stock,” also probably the German Biihne? which refers rather to the boards made out of the trunk of a tree; but the Greek word t9ouvdr ‘a bill, height, heap, mound,’ would seem to suggest an easier explanation of the Gaulish place- names ending in bonu. It has been hiuted by M. d’ilrbois No. 360. No. 451. 3 No. 651. 4 No. 2862. * It reminds one of Gartnait nitiperr, or Gartnaich Ditrberr of the Pictish Chronicle, whoso name is renderd in other chroniclev Oarnard Biuca and Garuad lc rick: gee Skene’s Chron. of the Picts and Scots, pp. 6, 28, 172. 200. If Diuperr or Diuberr be n loanword from a Brythonic dialect, it might be traced to the came root 88 the Welsh gwobr ‘a reward,’ for an wrly wo-pr or the like, cognate with gfw)o-brynu ‘to obtain by merit.’ while the verb dy-brpu ‘to obtain or acquire,’ suggestn n correlative noun with the mnto refix 88 Doquiw and hiuperr, in case they are to be analysed 88 here wumed. !n !n II note from Dr. Whitley Stokes I learn that he gives the preference to the spelling Diacpeir, as be connecta it with the Welsh pair ‘a cauldron,’ nnd invokes tbe parallel of an Irishman who was called ‘n cauldron of hospitality.’ 0 See Parthy 8 Pinder’s edition, No. 41G @. 197); in the lndex they identify Equabona with Couna, Copa. ’ See the fourth edition oi Kluge’s Dict. 108 P AND GROUPS. de Jubainville that the first part of Eqtmbonn is purely Latin. Of course, names like Juliobona and Augustobona existed in Gaulish, together with others into which the Latin proper names Augustus, Czsar, Claudius, Drums, Fliivius, and Julius entered; but at present I cannot recall an early instance involving a Latin appellative like eq1m. Eqirnesi and Quurquerni. These wero the names of two of the peoples who formed the so-called Conventus Bracar- ntrgustanus as enumerated in an inscription found at Aqua Flaviae. The peoples of this Conventus dwelt between the rivers Minho and Douro, where the principal tome were Bracara or Bracaraugusta, and Aqua Flaviz, now Braga and Chaves respectively. The Equaesi are SO called also by Pliny;a but the Quarquerni's name is given as Qiierquerni by him. Here Querquern- is probably identical with the Perpwn- of such personal names as the masculine Perperjia and the feminine Perpernia, both of which occur in Spain, namely at Tarragoua and Barcelona; also in Cisalpine Gaul, to wit, at Padua;5 but the most note- worthy instance, perhaps, is an inscription found at Martos, the ancient Tucci, not far from the eastern boundary of BEtica: it readsG-D . M . S I M . l'erperna Gallicanus I Annor. L . H . S . E . S . T . T . I, 1 Huic. Mer . Fil . Et . Xep . Fec I . Querqueru- and Perpern- look like redupliea- tione of II atem represented in Irish by crania7 and in Welsh by prenn, now written yren, 'a piece of timber, a tree,' which are presumed to be of the same origin as

The Berlin Corpus, ii. So. 2477. 2 Nat. Hist. iii. 28. lbid. 4 The Berlin Corpus, ii. Noa. 4301-2,4393, 4647, 4556. 5 Ibid. v. NOEL3004 and 2866. which is of doubtful reading. The name C. Pe mius. which occurs in a list of gladiators found at Venusiu in Apulia, proxbly belongs here : Bee the Corpus, ix. No. 466. Tbe relation between Perpcrna and Pcrpcnita is a question of considerable difficulty, which I am not prepared to discuss. No. 1709. 7 The form in earl Celtic must have been qynm of the neuter gender, and the reason why the rd uplication' should yield, not qyrqyrann-, but qrpqtpn-, is to be sought in tho too great accumulation of eonsonante the former would occasion. CELTS AND OTHER ARYAh’S.-J. RHTS. 109 the Latin quercus ‘an oak,’ quernus ‘oaken, of oak’: com- pare Santmnn, an old name of the Irish river Lee, Welsh Hafren ‘the Scvern,’ and Irish salnnn ‘salt,’ Welsh hakn, as illustrating the aame treatment of yn and In in Irish and Welsh respectively. Let us take next the other outlying portions of the Celtic world looking towards the sout,h, namely Gallia Cisalpina or Northern Italy and a portion of the Alps. Here we have a variety of names which may have belonged to a Celtic dialect; but several of them are too obscure to bo of use to us, such as Quas.suiuza,l also given as Cusonin,2 both in inscriptions at Verona; and such as Qcterras in another inscription at Verona. Then me have a people called Quadiates on the Cottian Arch at Susa : they belonged to the Cottian Alps, and were probably the same as the Quariates supposed to have left their name to Le Queyras in Switzerland. It is scarcely probable that they were the Germanic people of the Quadi of whom wo read in the Germania of Tacitus; far more likely is it that they were either Celts or Liguriana, if we may judge from the other peoples enumerated in the same inscriptions? Putting aside, then, such names as the foregoing, we find remaining in the inscriptions such forms as Equasia, Squillius, and Vciquasiiis. The stem of the name Equaaia found at Trieste6 might be either Latin or Celtican, but the formation of the word seems to resemble that of Vriquasiz~s,the Celtican origin of which is not improbable. This, together with Vequasiiia and Vequasia, occurs as a personal name in inscriptions 7 from Polenza and neighbouring places in Piedmont. The variety of the spelling shows that the rowel of the first syllable was long ; 80 that we may have here in fact the same atem as in the Irish namea Fiaciwa and Fiaciira. The genitive of the latter,

Berlin Corpus v. No. 3463. KO.3016. Y No. 3597. 4 No. 7231. See Mdlenhoff’8 Dsutache dlfcrtumskundc, rol. ii. 249. No. 591. 7 No8. 76948, 7680, 7682. 110 P AND Q GROUPS.

Fiachrach, appears in a late Ogam as Yeq!irecc,’ while that of the other, Fiuchna, is found in an Ogaln of the earlier sort as Veqoanai.‘ The later forms with their ia ehow that the original vowel was ci or e and the Irish common noun jach ‘ a raven ’ suggats a possible interpretation of this group of names. Lastly we come to the names SquiZZiua 9 and Sqiiillianzca in inscriptions at Verona; to these may be added SqueiZhniusJ from an inscription found at Narbonne, an8 poseibly a genitive Squeliolea6 in a Christian inecription at Maraeillee. The variety of spelling here also suggests that the vowel of the first sylhble was possibly long; and in that case the length mas probably the result of the elision of a consonant, the original stem being presumably squetZ of a neuter sq~etlh, meaning ‘ a piece of news, a tale.’ The Goidelic language of the insular Celts retained the t for centuries later, as we find on one of their monuments in Glamorgan the compound scitli- rissi, the genitive of an earlier sqtetlicisso- ; but the point of importance here is that the Brythonic treatment was quite different, seeing that the Welsh word for a story is chwedl, from an earlier s!,elZon, to which eqt+tldn had been levelled,’

1 The stone comes from Monataggart. in the parish of Donooghmore, in the county of Cork. and is now at the Royal Irish Academy * l’he stone is at Cooldorrihy. in the parish of Kilmichael, in the same coun . The o of Veqounni b unusual aod meant probably for the y (of the pi4 ordinan7 y represented by the single Ogam character I I I I I) : in another instance the I I I I I is followed hg the Ogam for u) or v, namely but the Ogrrm for q,u is nuvcr con- founded with that for e or k. Berh Corpus, I. No. 3335. No. 3401. 6 Berlin Corpus, sii. h’o. 5972. No. 401. ’ For other instances of 8~ for sqg see my Celtic Britain2, p. 92. The dis- carding of tho q waa here probably due to the syllable being unaccented ; so also in Welsh pcunydd ‘every day, daily ’ for qpwndii - from an earlier qpoqy-drj-. Compare the old Irish cefhv iour, the initial of which suggested the symbol I I I I for c in tho Ogam alpbabet; 60 the word wns probably at ono time cet;hes from an earlier qytpbes, corresponding to the Welsh peiguur, now pcdwur. This acccntuation is also that of the Sanskrit catv&a and it is impIicd by the d of GothicJiduor, and possibly by thc R of the Latin quutuor: see Mr. Wharton’s paper in the Phil. Tram. 1888-90, pp. 46-9. On the other hand. the former initial of the Irish ebic, cbig ‘five ’ 6uggesk.d the Ogam symbol I I I I I for q:‘, which the fifth numeral must have retained in Irish Gaelic : in fact Manx Gaelic CELT8 AND OTHER ARYANS.-J. RHYS. 111 and the same remark probably applies to Gaulish. Such a name then ae Squillius would mean a news-bringer or a scout, and perhape a story-teller. The ancient inscriptions of Illyricum present nothing very certain as regards the present question : we get a personal name Qctordaio* from the neighbourhood of Salzburg, but it is obscure to me as is also the genitive Quitrii,* which comee from KovAcsi, north-west of Buda-Pest in Hungary. It is probably more correct likewise to regard q names, such as Q~ert/io,~from the Roman town of Aquileia and its vicinity, a8 Latin rather than Celtic. So we now recross the AIps to Transalpine Gaul, where we have already found the name Squeillnnius and others in point. A word must now be said as to the relation between the two Celtic pcwples of the Continent, and first from the geographical point of view. If we take for granted, as me eafely may, that the earliest population of Gaul was not Celtic or even Aryan, and further that the Celts of ancient Gaul did not originally corm thither from the direction of Spain or of Italy, we are entitled to conclude that the Q Celts arrived in the west before the Y Celts, as they are found occupying the furthest parts of the Celtic area, namely north Italy, Spain, southern and western Gaul, Ireland, Man, the Scotch Highlands and Islands. Tho coiiclusion is scarcely to be avoided that the later comers, the P Celts, came as invaders and conquerors constituting themselves the ruling people wherever they could vanquish the other race. This seems to have been the case in most of Gaul, and some of the results of this state of things are indicated in CEesar’s Commentaries,* as, for instance, still rotains it in its queig ‘ five ’ 88 contrasted with kegcesA (pronounced kcgfrh or b&h) ‘a fortnight,’ in Irish ciiedigia or edicthighis, with 6i owing to the analogy of dig. 1 See the Berlin Corpus, iii. No. 6.523, from which it appears that Conginna, daughter of Quordaio, had married the grandmn of a man bearing the name Afrvul-i, which looks like Celtic. 2 No. 3621, but Quita’a daughter bore the apparently Celtic name of Comaftmora. 3 Vol. v. 30.1270. ’ Booka i. 4, vi. 13. 112 P AND Q GROUPS. in the muItitudes of clients and dependents forced to attach themselves to the Gaulish chiefs. This would also explain the scarcity of data as to the language of the earlier Aryans there; for Roman and Greek writers would come more directly in contact with the ruling race and hear Celtic names only in the Gaulish form which the ruling Gauls gave them; and, to judge from similar cases elsewhere, the subjugated race must have acquired at an early date the habit of translating its own names and forms into those of the conqueror’s idiom. We seem to have an instance of this in an ex-voto to certain Nymphne Augustae in the neighbour- hood of Vaison in GaIlia Narbonensis, where the name applied to them is Percernes,’ in which we seem to have an imperfect translation into Gaulish of the qverqgers- of the proper name Querquerni ; 00 the epithet may be supposed to have characterized them a8 nymphs of the forest or wood- lands. But be this as it may, that is the sort of process always going on in dialects brought into contact with a dominant dialect. Nevertheless there are one or two well-known names which for some reason or other successfully held their own, and those are Sequana and Sequani, and Aquitani. How the name of a river like the Seine should have been accepted by the invading race is not very difficult to understand, but the difficulty is greater when we come to that of the people known as the Sequani, as there are no data to decide whether they were Celticans or Gauls. Their geographical position, however, among the mountains of Switzerland, and the r6Ze played by them in Gaul at the time of Czsar’s advent and for some time previously, are not inconsistent with the possibility of their belonging wholly or in great part to the earlier race. Cacsar represents3 the Bdui and the Celticau people of the Arverni as the heads of two rival federations;

1 Vol. xii. No. 1329. It occurs d80 88 a woman’B name, in an inscription from the neighbourhd of Trevisio to the north of Venice : 8ee the Corpus, v. KO.2129. Most prohably S?quana wan in the Amt instance tbe name of a goddess (see Rev. Celt. vol. Iii. p. 306) ; but a compound name Srquanuioluoa occur6 ou coius abcribed to the Bequadi. f Bk. i. 31. J? AND Q QROUPS. 113 and that of the Arverni proving the weaker, it was they who, together with the Sequani, invited Ariovistus and his Germans to come into Gaul. The presumption is that both Squana and Sequani are words of Celtic origin, but with regard to the name of the Aquitani we are left in doubt whether it was Celtic at all or not; if the word is Celtic, it may be supposed to involve a Celtican word for water, of the aame origin as the Latin qua, and that the first bearers of the name Aquifani mere in that caae described as a people dwelling near the water, meaning the sea, and this would agree well enough with the little that is known of the history of the word. Strabo 1 maintains, and probably with perfect justice, that the people beyond the Garonne, whom Cacsar called Aquitani, were more Iberiau than Celtic. From them the province of Aquitania, constituted by Augustus so aa to extend to the Loire, took its name. Pliny makes the Iberian and original Aquitania into Aquitanica, adding, in the passsge already cited, that it was at one time called Aremorica. So it is possible that we have here to do with a torm Aquitan or Aquitanic, admitting of being rendered into Oaulish by the adjective Ar’enioric, which meant ‘ maritime, belonging to the sea-coast.’ The former may in that case have been derived from a Celtican word of the origin already sug- gested. Thus Pliny’s Armoric Aquitanica coincides with the information which led Caesar to confine the name Aquitani to the on the further side of the Garonne. Pliny traces the name there to a single tribe bearing no other deaignation than that of Aquitani : their exact position is not given, 80 that one is at liberty to suppose that they dwelt on the coast somewhere between the Garonne and the . Though originally synonymous, the later usage had the effect of severing the terms Arernoric and Apztitanic. Areiiioric or Armoric became associated with the Armoric League, and shifted with the shifting fortunes of the states constituting it, which Caesar describes as civitates qua!

See Meineke’s ed. iv. 1. I. Phil Tram. 1891-28. 8 114 CELTS AND OTHER ARYAKS.-J. RHYS. Armoricce appellantur.’ On the other hand, the other adjective was associated by Augustus with the province of Aquitania, including the spacious region from the Garonne to the Loire; and it is there that the name Aquitania has ever eince had its home, and has come down to later times in the French form of La Quienne. Here must be mentioned the term Chortmicum, which occurs, among other geographical names, in the wessobrunn Codex in Munich, a manuscript written before the year 814. The names of CeItic interest in it are the following : Hybernia . ecottono laid, i.e. the Land of the Scots. Qnllia . uualho hnt, i.e. the Land of the Welsh. Chortonicum . auh uitalho Zant, i.e. also the Land of the Welsh. Equitnnia . uzca~conolant, i.e. the Land of tbe Vnecones or Gascons. Dom- noniam .prettmu, latit, i.e. the Land of the Britons. But the name of prime importance here is Ckorlotiicurn, a spelling which may be regarded at3 standing for Cortonicum in the text which the writer of the Old High German glosses had before him, and Pott with his usual insight perceived that it was to be interpreted by means of the words Cruithne, Cruithneach, ‘a Pict.’* Cortonicum would thus have to be

Bks. V. 43, Vii. 76. In both paasagea the spelling ia drmorieos. nor does Holder mention any variant like Awnoncue, which, etymologically speaking, would be the older and better form of the word Prof. Ncttleship rcminds me that the prevalent form in the poets is scanned Ar--Ns. a Etym. Forscb. 11. ii. 899 : sea also Windiach in Ersch & Gruber’s Ency- clopredis, B.V. EeNUche rprudm. 1 copy the gloaeer, from Graffa Diutiska, vol. ii. pp. 370-1, in the hope chiefly that some one will inform me to what text they belong : it is needless to point out the curious queatiom which some of them 8Ug-t :- Fol. 610. Hec nomina de rarib prouinciis. Hpbernin . acottono lant. ,, 616. Gallia. uualbo Iant. chortonicum . auh uualho lant. Equitania . numono lant. Uacea . uuaacun. Germania. franchono lant. Italia . lancparto lant Ausonia . aoh lancparto lant. Damnonism . prettono Iant. Brutmi. prezzun. Arauea . sarci. Ispania . benauentono lant. Cyunari . suapa. Pannonia . sic nominatur illa hameridip danobia . et uuandoli habent hoc. P AND Q QROUPS. 115 explained as an adjective formed, probably in Latin, from the name of a people called in Celtic Qytiilones or Qptlto>/!i, of the same origin not only aa Cruithne ' a Pict,' but also aa the Welsh P,ydyn, the part of Britain inhabited by the Picts of the North, and Welsh Prydain, the name of the whole island. So Cortonicum may have meant all Qaul, or a part of Gaul, or else it may very probably have been u~edin both senses ; and to discover approximately the part of Gaul it could have apecially referred to we cannot do better than take the synonymity of Cruitkne and Pict in the British Islea aa our guide on the Continent. This clearly SUggCIh3 that the Cortonic district par excclhce covered the whole or part of Poitou or the land of the Pictonea, whose name, together with that of their town now called Poitiers, cannot, etymologically speaking, be severed from the name of the insular Picte. Thus it may be presumed that the term Cortonicum referred to the region so called, whether all Gaul or only a part of Gaul, as the land of the non-Aryan aborigines ; and it goes to show that the early Celts regarded the latter as one and the aame race, whether on the Continent or in the British Isles. That ia too large a subject, however,

[Sote continued from previous page.] Fol. 620. Amoricus . pigiro lant. Iehie . peigtra . Ister . daoobia Sclauus et auarus . huni et uuinida. Paleatins . iudeono lant . hoe wt circa hierosolima. Cuandnli . huni . et citta . aut uuanddi . Auriliana . sic nominatur illa terra ubi roma stetit. Pentapoli . sic nominatur illia patria . ubi rapana stat. Tharcia . illa patria . ubi constautinopoli . stetit. Cynocephali . canini capita. Amazon- . hoc .s. uirgiues. ,, 626. Thehaida . illa patria inde foit mauricius . Argi . greci. Ethiopia. patria maun . De Ciuitatibus. Luctuna . Iiutona. Argentoratensb . 8trazpunIC. Nimitcnsis ciuitas . spira. Uuangiaonium . ciuitas uuomacip. Agippioa . choloone . Constantinopoli . coetantinoses puruc. Neapolis . ciuitas noua. Norica . rc anespuruc. Allotla . Auy6nsa. Betfagia . pazauua. ,, 63a. Ualuicula salrpuruc. 116 CELTS AND OTHER ARYANS.-J. HHYS. to be discussed here; let it suffice for the present to have called attention to the form of the word Cortonicum, as it proves to be derived from a Celtican source, not from a Gaulish one. Its history, nevertheless, is incomplete ; for it is not known in what author it was found by the writer of the glosses in the ninth century ; 80 one cannot my whether the text he read is still extant, or on the other hand whether it did not carry him back a con- siderable distance of time towards, let us say, the fourth century. I say the fourth century, as we have evidence in the dialogues of Sulpicius Severus that not only Qaulish but Celtican likewise was in use in his time : his words are, Tic cero, inquit Poskminws, tel Cellice aut, si mauis, Gnllice loqtcewi dzirnmodo dlnrfiuztni loquaris.1 The words which Sulpicius puts into the mouth of Postumianus deserve all tbc more attention as Sulpicius is said to have belonged to an influential family in Aquitania, where Celtican speech may not unnaturally be supposed to have been last heard. The word Prydain, which is the Welsh name for Great Britain, is not etymologically related to the latter name from the Latin Brifatinia, which was formed from the Latin name of the people whom the Romans at first called Britmtti. For after the Romans conquered a part of Britain and became better acquainted with its peoples the term Britami mems to have gradually fallen into disuse in favour of the kindred form BrittGnes, which is also the pronunciation of the name to which the Rrythonic Celts bear testimony by their use of such forms as Welsh Brytlton ‘ a Briton or Welshman,’ Brylhoitiy the Brythonic language, Welsh,’ Old Cornish Rwfltonec ‘ tbe Erythonic language, Cornish,’ Breton Brczotzek ‘ the Breton language.’ Where then did the Romans learn their first name for the Brgthons? Probably in the Iioman province in the South of Gaul. At any rate it is remarkable that the old Irish word for Bryfhons, which had also to do duty for Britain, was the plural Rretuin, genitive Rretnn, dative Brettiaib, accusative Bretnu, which corresponds to the

Halm’s ed. Dial. i. 27, 4. P AND Q GROUPS. 117

Latin Bjsitanni, or, more exactly, Brittniii. Hence it would appear that the Romans must have learned the name from a people who spoke a language resembling Goidelic, that is to say from the Celticans of Southern Gaul or the Roman province. This is corroborated by the fact that Britand and Briltanni resemble the Qreek form Bpmauol, which reached Greece probably through the Greeks of Mar- seilles, who doubtless learned the name in the aame district as the Romaus : perhaps it would be more accurate to regard the Britanni of the Romans merely as the Latin rendcring of the Greek Bpen-auol, and to suppose the Greeks to have Iearnt the latter from the Celts nearest to Marseilles. At nl1 events it was not the form which the Brythons themselves gave to their own national name. It probably represented rather the pronunciation which the Q Celts of the Continent gave it, as it must hare done, with tolerable accuracy, that of the Q Celts of Ireland, judging from the forms which the name exhibits among them in later times. Such are the traces on the Continent of a Celtic language resembling Irish as contrasted with Welsh, and though they seem to me to render its former existence there highly probable, I leave othere to settle how far it is probable.

11.

The question of classifying the Celts into two groups characterized by the use of P und Q languages respectively is not to be dismissed without reference to a similar grouping of the Aryans of Italy and Greece. Thus to begin with the former, the Romans used qy just as the ancient Irish did, but in so doing they stood well-nigh alone in the Italy of historical times ; for the Osco-Umbrian dialects replaced 9F by P-'

1 Nessapian and possibly other dialects of the south are to be regarded BR more akin to Greek than to the other languagen of Italy : see Mornmaen's 77tih- ltdmh Diatckk, p. 86. 118 CELTS AXD OTHER ARYANS.-J. RHYS.

The Latin-speaking Italians of the beginning of Roman history occupied a comparatively small area, their original territory in the peninsula having probably been narrowed ' by Umbrians and Sabines, by the latter of whom the legends of Rome represent the nascent city hard pressed. On the other hand the ecanty remains of some of the dialects of ancient Sicily are supposed to ehow close similarity to Latin, and this euggests that Latin or kindred dialects were once spoken over the whole of the eastern coast of the Penineula, from Latium to the Straita of Messina. Their disappearance from all that tract waa probably due to the conquests made within hiatorical times by the Samnites, and other tribes of Oscan stock, and in part also to the Greek colonies planted on the coast. In the rear of the Oscane came the Umbrians, occupying not only the district which bore their name in Roman history, from the Adriatic to the Apennines and the Tiber. but also the adjacent country westwards to the sea opposite Corsica. In other worh they owned, if not the whole of what came later to be Etruria, the south at any rate and the east of Etruria, 80 that they completely barred the north of Italy from eea to sea behind the peoples of the Siculo-Latin stock. I cannot discuss here the extent to which the latter had been deprived of their territory by the Oeco-Umbriane, or the fact of these last being in their turn deprived of so much of their territory by the Etruscane, that Pliny found the Etruscans to have taken possession of no less than 300 towns of the Umbrians. The one thing to be borne in mind for the present is,

' The remains of the Faliscan dialect ahow clearly that tho people of Fderii, for inatance, belonged to the name stock an the Lath and not to the Umbrians or Ossans: see Deecke's 'Falieker' (Strassburg, l888), where, pp. 135, 15ti, 193, he give8 an Faliecan -put or -m=Latin que, cuondo=Latin quado, and other simdar evidence which cannot be overthrown by the Ocenrrence in a Faliscan inscription of Puponio equated by him with Pmponiw. at the name time that he snggeak equating Popw with hbh: we pp. 163, 187. 2 See ifi. 112: the ie highly instructive, especially the following words-" Ab Ancona Gaiica om incipit togatae Galliae cognomine. Siculi at Liburni plurima eins tractua tenuere, in primie Palmenaeni, Praetutianum Hadrianumque agrnm. Umbri eon expulere, hos Etruria, hanc Galli. Uni- brorwn gens antiqukima Italiae exbtimatur, ut quoa Ombrios a Graecis putent dictos, uod inundutione terrarum imbrihus superfuiseent. Treeenta eorum oppida '&mi debellam reperiuntur." P AND Q GROUPS. 119 that the Italians of the Q group appear to have entered Italy betore those of the P group, that the Siculo-Latin race had alresdy settled down when the Om-Umbrians arrived. Much the same kind of remarks may be applied to Greece, where the Q group is most obviously represented by the dialect of Herodotus,’ with such forms as K&)F and K~TEfor the .rr& and T~TEof $he other dialecta. The historian was a native of Halicarnassus in Asia Minor, and his Greek belongs to the group of dialects called Ionic. As there is no reason to suppose that he invented the language of his writings, he must be supposed to have heard it at Halicarnassue or in some part of Asia Minor, or else, as has been eometimes thought more probable, in the island of Samoa, where he ie said to have passed eome years of his life. Now Halicarnaesue is spoken of as a Doric colony, namely, of Troezen; but this by no means proves that it was not also Ionic. That it was 80 is seen from the fuct that the Ionian deity Poeeidon was worshipped at Halicarnassus with peculiar zeal,P and the connexion itself of Halicarnaesus with Troezen is by no means in- consistent with this view, but rather the contrary; for Troezen mas a mixed city. Originally Ionic, it had been conquered by Dorians. The name Ionian is applied eaily to the inhabitants of the north-east of the Peloponnese, including Troezen and other places; also to the district 011 the north coast, subsequently known as Achaia. In what numerical proportions the Dorians and the Ionians Iived in tbe towns conquered by the former, it is impossible to say ; but in Sicyon, for instance, the old inhabitants formed a fourth tribe, whereas at Corinth they consisted of five tribes, while

1 Herodotus ia selected an the representative of the dialeet, for one learns from the editione of such authors 86 Iiippocratee and Heraclitas t.hat the manuecriptu nar~owthe domain of the I[ ns compared with the uaage of the bistorian. It wan doubtlesa very natural for Greek ecribea nnd editors to normalize the dialect by giving its fnrms the T with which tbey were familiar in Attic prow; but the longer the writings of an author were, the more resistance they offered to this prom of mimilation. 2 Evelyn Abbott’n Greece, vol. i. 12.5 : Miiller’a Dorier, i. 407, ii. 106 ; and the Berlin Corpus of Greek Inscriptions, No. 2655. 120 CELTS AND OTHER ARYANS.-J. RHYS. the conquering Dorians only made up the three remaining tribes. Tho conquest of the Peloponnese known as the return of the Heraclidac was effected by Dorians helped by Btolians, and the former can he traced back to a district between Oeta and Parnassus, where their name survived in Doris: at a still earlier period they dwelt in the Thessalian district of HestiZotis.1 When the Ilorians and their allies conquered the Peloponnese, the ancient Ionians and kindred peoples were subjugated or driven into corners, such as the north-east of the peninsula, together with Attica, also Euhoea, the other islands and Asia Minor, whither the Dorians in due time followed them. Eventually ensued a fusion of races, which it would be useless to attempt to analyse. But it will suffice for the present to bear in mind that the Dorians and kindred invaders found the Ionians in the Peloponnese before them, and that t.he Ionian dialects contained among their number one, at least, which belonged to the Q group. The principal questions which these remarks raise may be comprehended under two headings, the fusion of P and Q Aryans and the common descent of the I? Aryans. Let us take the latter first: is the modification of qu into p enough to prove the nations speaking languages of the P group to have Tread from a single centre, whether they are found in Celtic lands, in Italy, or in Greece? As the physiological change here in question might, ao far as one can see, take place in any language having words with qv, that change would be a precarious foundation for such a conclusion.z But it makes a very great difference to find that the change haa been resisted in the outer area alone, that is, among the Cloidelo-Celtican group in Celtic lands,

1 Abbott. rol. i. 61 ; Herodotus i. 56, viii. 43. * Most of those who notice this, spkas if they consider the chan e of qu into p a very cnmmon phenomenon, wherean. besides the cam of the Aqan!anguagen in question, I have not succeeded in eliciting an instance from hen& acquainted with non-Aryan languagee. One is nerally asked to take Roumanian LIB in point; but that will not do ae an in%pendent instance. For who h to my that the legionary ancestor of the mcdern Roumanian who aays patru for the Latin grmfwr, WBB not an Oscan or Umbrian, or even a Gaul, inheriting the p pronunciation ? P AND Q GROUPS. 121 the Siculo- in Italy and some of the Ionians among the Greeks. In fact we have only to make a simple calculation of chances, which may be put thus: let the three P groupe of the inner area bo represented by three white balls, and the three Q groups of the outer area by three black balls ; then suppose the six balls placed in a bag, and the question is, what the chances are, in drawing three at a time, that thoee three will be the white balls. Now the number of different triplete that can be formed of six balls is :%;=20 : that ie to say the chance of drawing any given triplet is &, or, in other words, the odds against it are 19 to 1. Suppose we Ieave out Greece on account of the evidence being lese clear there than in the case of Italy and Celtic lands, our calculation is then made with four balls instead of six, and the number of pairs posaible is g:=6 : that is, the chance of drawing any given pair ie 6; in other words the odde against it are 5 to 1, which is a preponderance of probability by no means to be despised, and from which one might proceed to assume the same state of things in Greece as in the other two linguistic regions. It thus appears unreaeonable to suppose the change of qv into p among the Aryans of the iuner area to have been a mere matter of accident, and the alternative explanation which aloue deserves considera- tion is, that the P Aryans iseued from a common centre where the phonetic change in question had been accom- plished once for all in their common dialect. Here may be added the fact that the consonantal change is not the only characteristic of the P languages, as they Bhow another change, namely that of long u into long ii or even i; but aa this will come under notice later, Iet it s&ce at this point to remark that it vaatly increasee the etrength of the argument.'

1 For instead of aim Ip dmwing the three white balls once, the chance of which taking place waa stown to be only gST,one hna to suppose them (replnced and) successfully drawn a second time. The chance of the compound event is &x&J=~~:that is, the odds asinst it are 399 to 1. If Greece be omitted the chance of the corn ound elcot is 8x6 =J6, or 35 to 1 against it so far 88 concern the Celtic and)Italian are,. 122 CELTS AND OTHER ARYANS,-J. RHYS.

Irish, Latin and the Qreek dialect in point’must so far a8 concerns the consonantal chauge which interests us here be regarded a8 having remained in a senee on the level of Teutonic, ae a tringle instance will suffice to show: take, for example, the English interrogative who, Gothic hm, and compare with it the initial of the Irish cia ‘who,’ Latin qui and pub, and the Herodotean KO~OF,contrasting with them the Weleh pwy, Oscan pie ‘who,’ and the common Greek ITO~OOF. Be no philologist supposes the change to have been fromp to qu, Irish, Latin and Herodotean Qreek must be regarded as having in thie reapect remained on the ancient level, the reduction of q? top being a later phenomenon. A similar remark applies to the vowel change to which alIusion haa been just made. So the two changes agree in sigualizing for us the rise of a new Aryan dialect, and this translafed into ethnology can hardly have any other meaning than that Aryan speech had been adopted by a non-Aryan race, which introduced into ita adopted language the habits of pronun- ciation prevalent in its previous language. The question next comes, where the conquest of non-Aryans by Aryans took place, and looking at the distribution of the Aryans of thie mixed origin as represented by the P dialects of Gaul and Britain, of Italy and Greece, also poesibly by tbe languages of peoples of the Thraco-Phrygian stock both in Europe and Asia Minor, one cannot help thinking that it must have been somewhere in the region of the Alps. Granted this, we are enabled to explain a variety of the features which characterize Celtic, Italian and Hellenic dia- lects : in other words, we now come to the other question to be noticed, that of the fusion of the later or P Aryans with the earlier or Q Aryans. But here, at3 before, I only attempt to deal with the race9 indirectly, through the linguistic data which they have left us, aud Latin proves of prime importance to the inquiry. The history of that language is a very remarkable one, for, after having its domain narrowed to a comparatively small area, it begins to conquer all the dialect8 around it, nor desists till it becomes one of the great languagee of civilization; but what ie particularly to be P AKD Q GROUPS. 123 borne in mind is the fact, that the antagonism between the makers of Rome and the peoples around them seems to have been BO intense that Latin remained comparatively unaffected by non-latin diaIects up to the time of its classical period. Thus not only does Latin agree with the rest of the languages characterized by qi! in the matter of that combination, but it further abstains from the correaponding change of g? to b, which happens in Irish and Greek. It retain8 gu or reduce8 it to o: take, for examplo, tho word unguo ‘I Emear or anoint,’ unguen ‘fat, ointment,’ Oman umen (for ‘unrben), Irish ittrb, genitive imbe (like ainni ‘name,’ genitive anme), Breton aniann, Welsh ynenyn ‘butter.’ Take also the Latin orburl ‘living’ (for *ggiwos), Oscan bivus “vivi,” Irish beo, Welsh byw ‘quick, alive.’ A remarkable exception in Latin ie bou ‘ an ox,’ which, had it been native Latin, ought to have had eome euch form as cos, and not Los; but the very fact of the latter standing pretty uearly alone speaks clearly for the comparative freedom of Latin from any considerable admixture of Oscan.’ Another reepect in which Latin remained pretty nearly on the old level is that of long ‘u, which in the Y languages shows a narrow- ing towarde French ti or even f. Thus while Latin had 81(S ‘ a sow,’ Greek had its &, pronounced with a vowel like the French ti of the present day; and the accusative singular and plural of this same word were in Umbrian eim and $if respectively. Umbrian had also pir andfrs corresponding to the Greek &p ‘ fire,’ and Latin fruges ‘ cropa.’ Similarly Latin tir, English thou, appear in Irish as ta, but in Welsh aa ti, and Irish cir ‘a hound,’ is in Weleh ci, not to mention that such a word as Irish luath, in Baulish touta ‘ a tribe or people,’ is tid in Breton, and in Weleh ttit and ttid, with a u which in North Wales ie much narrower than French ti, while in the rest of the Principality it becomee moatly i. This may be regarded ae a Brythonic extension of the domain of the tendency to narrow or unround the vowel. The comparative freedom of Latin from the changes here in question suggesta the conchion that they wore phonetic

1 If b08 be not Oncan or Umbrian, it may be of Gauliah origin, or else a loan- word from Greek, which some are inclined to maiutain. 121 CELTS AND OTHER ARYANS.-J. RHYS. characteristics of the P dialects, that is to say, all three of them, the labializing of q? into p, of gu into 6,‘ and the modifying of u towards 2.2 Granted this, it follows that one must suppose the linguistic influence of the later Aryan comers to have varied condderably in force a8 measured by the changes it effected in the phonetics of the 9. dialects of t.he Celts, the Italians and the Greeks respectively. In Italy, as already stated, it may be said to haw been a minimum quantity, but it waa quite otherwise among the Celts and the Greeks. Let us begin with the latter: the majority of the Greek dialects are known to have used T for tho qv of the labalizing Aryans, but they did not treat that combination always in the aame way. Before the vowel o they made it into T ae in aoZ9 ‘ qualis,’ but before e and t it became r, aa in -78 and dq, with which are to bo contrasted Latin -pie and pi8, while before or after v tho guttural remained, as in K~AOP‘a circle,’ A.-Saxon hlceohl, English whed Under similar conditions 911 became B, 6 and y ; aIso gh? became +, 8 and x respectively. No such limitations of the phenomenon of the guttural plus the y yielding a labial are known to have systematically existed in the P dialects of either Italy or the Celtic world,s and certain Greek dialects seem to have somewhat defied them, as

The Change of ght intof or q~haa been omitted, partly as being of minor importance here, and partly aa coutained, 80 far a8 concerm Celtic, under that of g? into b. * My,friend Mr. Wharton, in the paper to which reference has already been made, given it aa his opinion, p. 64, that “ the influence of dialect on the Homun long vowels, and ospeciillly on the diphthniilp. was much stron er than on the short vowels, as conversely that of accent wus much weaker ;1’ But from among his instances of ti becnming i, I should he incliiied to remove such aa divinus, amicus, formido, vopiscus, lls contrasted with opprtiinue. cadficue. testiido, coriiwus. and the like; for the use of a termination inus, for example, rather than rinus, raises questions aa to the thematic vowel, and it is open to the suspicion of being due to analogy rather than to a mere phonetic change. Among Wr. Wharton’n other instances may possibly be 60me which were due to the influence of Oscan and Umbrian. But Prof. Zimrner in his XeZfi8cb Stdim in Kuhn’a Zeiteehrift, vol. xxx. pp. 134, 140, regards Irish guth ‘voice’ and Greek &fj as of the 8ame origin, md treats gcgon 6 I have wounded,’ aa the perfect of a verb which in ita other tenses baa b, such 88 him‘ I strike.’ So here we have g? becoming gu or go, and some such a rule must have applied in the case of Sequann. supposing it to be-against the robabilities of the case-Gaulish and not Celtican. Add the form SEQVONIh3, which, if correctly read, ia deserving of notice iu this con- nexion : ixe the Xcuus Ctlliquc, vol. ~IL p. 307. P AND Q GROUPS. 125 when the Theswlians said KI~for 719 ‘who.’ Then as to the dialects characterized by the unrounding of long u into long ii, similar treatment is found to have been extended to short U, which was changed in the same dialects into ii ; and this still remains the sound of v in certain dialects of Greek, such as the Tzakonian. But the ancient dialects retaining u and 2i are said to have been Baotian and Laco- nian, probably also Arcadian, Cypriote, Pamphylisn, Chal- cidian, and Lesbian ; and such instances may be mentioned as the Bceotian name E6Boupos for the Attic EiXfipc, and the Laconian verb povul88~~for what mould have been in Attic piK& in the sense of XaXei ‘speaks.’ On the whole the P dialects may be said to have had the upper hand in the Greek world, and the fusion, not to say the confusion, of the dialects of the two groups must here be regarded as of such a nature that Greece cannot help us much in the attempt to distinguieh between peoples of the P and Q groups: rather have we to assume that tho conclusions already drawn as to the corresponding peoples in Celtic and Italian lands may be applied in principlo to Greece likewise. On Celtic ground we have no data to enable us to judge with any precieion of the attitude of Gaulish and Celtican towards one another on the Continent, but we seem to have a distinct trace of the influence of the former in the French pronunciation of Latin long id, a pronunciatiori which not only characterizes the French language, but is found aleo in modern Provencal, in the Engadine, and in Lombardy.2 Laving the Continent, wo have in those islands the facts of Brythonic and Goidelic pronunciations to draw upon. In the firet place it may be mentioned that the modifying of ir towards u or i is unknown in the Goidelic dialects, as is also the reduction of qy to p ; but, on the other hand, b for g? and ghy ie as general a rule in them as in the Brythonic dialects. How then ie this to be explained? To some extent, probably, by the geography of the British

1 Brugmann. 4i 48, 66. Dies, Gram. &r romuniscAcn &ruch (BoM, 1870), POL i. 42G. 126 CELTS AXJI OTHER ARYANS.-3. RHYS.

Isles ; for the case would be met by supposing the a Aryan0 to have first conquered Britain as far,'sooner or later, as the sea separating it from the sister island, and the P Aryane to have eventually arrived in such force as to establish themselves in the east and south-east of Britain, preserving their own language at first unmixed, while influencing that of the Q Aryans to the west of them more and more as time went on, and reducing their territory. In fact we may suppose the same thing to have taken phe here as in Qaul, namely, that the P conquerore everywhere made themselves the ruling class, even where the Q dialects conhued to be spoken. It was the speakers of these mixed dialecta, that is to say, of Q dialect8 modified under the influence of those of the P group, that probably sent forth men to conquer Ireland; they set sail, let UB say, from Angleseg and landed on the opposite coast, somewhere between Bray and Dundalk. In Ireland the Q language of the invaders became stereotyped against the further influence of the idioms of the P group, while the Q dialects left behind in Britain continued subject to that influence until they were assimilated out of all separate existence. It is needless to say that whilo Brythonic made this sort of conquest, it could not itself escape being phonetically modified, namely, by the previous habits, of pronunciation of those who adopted it inetead of their own language. Add to this the subsequent influence of the extensive conquests made in Wales and Dumnonia by invaders from Ireland about the end of the Rornaii Occupation. Such at any rate would be a possible account of the rise of Goidelic as it was when it spread from Ireland to Man and Scotland; but with regard to the infiuence of the P dialects on it, only one instance haa been specified, namely, the change of g? into b. This must, however, not be regarded a8 standing alone : there ie probably one other which must be supposed to date quite as early: it belongs to the history of the treatment of Aryan p in Celtic, which must by no means be confounded with the treatment of Aryan k or 99 whether changed into p or not. Thus it has P AND Q QROUF’S. 127 been for some time known to phiIo1ogist.a that Aryan p wholly dieappears in the CeItic tongues, that is, it has either given rise to another consonant or it has simply left its place empty. Take for example the Irish word Mr, Welsh thaw, which means $om and is of the eame origin, or the Old Irish word othir, which means futher and ie in fact the same word with fither and Latin pater, with their con- genera ; similarly the Irish verb lenim ‘ I adhere, follow,’ reduplicate perfect (3rd singular) liZ, (3rd plural) ZeZtar, future lilit ‘ sequentur,’ Welsh dilyn, canlyn ‘ to follow,’ and erlyn ‘to prosecute,’ are of the same origin as the Lithuanian limpu ‘ I adhere,’ preterite lipazi, Old Bulgarian prifinp (for pril+ttp) ‘I cleave to or remain hanging.’ 1 This consonant has disappeared even in Gaulish, aa may be seen from euch words as Aremoricus, where the prefix are, Welsh ar ‘on, upon,’ is of the same origin as the Greek rap& and the whole adjective may be rendered rapa9akiaucoy ; take aleo the name of the Belgic people of the Remi, who have left it to the town of Rheims, the ancient nurocortorum Hemorurn, where REnii is of the same origin as the Welsh rhzr?Cf ‘a king or ruler,’ Latin primus: compare English Jr6t and German furst ‘a prince,’ and the Latin prhceps, whence, through French, the English word prince. In other Celtic words the original p is represented by a guttural, as in the Irish secht ‘seven,’ Welsh saith, seith for ‘secht, or the Irish word umat ‘ noble, high-born,’ Weleh ucheE ‘high,’ which waa in Qaulish iicael- as in Uzellodunum meaning High Town. Thie adjective is parallel to the Weleh isel ‘low,’ and the two are to be traced back to the prepositions which have yielded the Welsh yn (Irish i n-, i) ‘in, into’ and *upI2 of the same origin as the Greek word i+,A& ‘high’ and English over, German iiber, Sanskrit upari, ‘ over.’ In all

1 Bmpannn, i. 4 36. 2 Thia etymology waa snggetlted .some yearn ago in my Celtic Britain? p. 310. but an older etymology is accepted, among others, by Drugmanu, who, in hrs (?rundrirr dcr urrgl. (Jmmmnfik, vol. i. 434, connects Urello- mth Greek oGEw I increase,’ and Lith. drkmfats ‘ high.’ 1P this were correct, one would rather expect the word to have begun in Gaulish with ou, cu, or ou, which however it dwnot. 128 CELTS AND OTHER ARYANS.-J. RHYS. these instances it would suffice, phonologically speaking, to regard p as having in tho first instance become h (in the Celtic dialects with p for qv), just as the samep is known to have done in Armenian in such words as haL ‘ father,’ the etpo- logical equivalent of that word and of the Irish athir, Latin pater, etc. H before vowels would disappear on the way down to the Neoceltic tongues, while before t it may be supposed to haw helped to produce the favourite combination Xt, retained in Irish, but modified iu Welsh to ilh, as in Old Irish seclrt ‘seven,’ Old Welsh seith, now saith. Before s the analogous cornbination xs was liable to undergo the changes illustrated by Irish uasal and Welsh uchet nlretldy cited .I The theory here propounded leaves the Goidels nearly related to the ancient Latins in harmony with the striking similarities between the Irish and Latin languages, at the same time that it represents the same Goidels as inseparable from the Brythons by reason of manifold mixing. So me should still be justificd in speaking of these two peoples aa Celtic, and not merely as Goidelic and Brythonic, or as groups only distantly related to one another, which is the utmost one could have said of them when they first Carrie in contact with one another. Tho same kind of remark applies to the two sets of Aryans in Italy and in Greece. But who were these races di8tinguished as those speaking Q and P languages respectively? The former, suffice it for the present to repeat, were of the 681110 Aryan stock as tho Teutons, Slavs, Albanians, Armenians, Iranians and Hindus, that is to say in so far as those nations were not merely non-

The case of initial sp making in Welshf (writtenf) and in Irish a has been omitted here as probably representing an original Aryan sp-h or ap aspimted : compare Welsh fcr ‘ thc ankle ’ with the Greek ccpupdv of the same meaning, and Welsh yon ‘ n staff,’ Jon dnfr ‘ a sling,’ Irish amn, with the Greek 09~~66y ‘a sling.’ In such Irish words as sonu the a hns probably replaced n previous f, as is the mein such borrowed words a~ Irish wion =Latinfrhrum, and siist= J,atin fustis. Some more instances of initial f in Welsh will be found brought together by mein the Rw. Cell. vol. ii. pp. 335-7. Sp-A or sp‘became cporfin both Brythonic and Goidelic, and in Brythonic thefhaa remained while it hm become L in Goidehc. In nny caw I see no room for the m, which Dr. Whitley Stokes suggeats in his Celtic Deckmion, p. 26. P AND Q GROUPS. 129

Aryans Aryanized by Aryan conquest. In Italy they appear to have been not only the Latins of history, but the race more widely represented by an early civilization, of which tracee occur from Etruria to Sicily, aa ehown by recent archsological reewrch. In Greece they were poseibly the rnysterioue people called the Pelaegi with whom Herodotus identified the Ioniane. But the Pelasgi, though Aryane, were not exactly Greeks, however readily they became Cfreeke in the sense of loeing their national identity in the ganglion of peoples which went to constitute the Hellenio world of hietory. In awwer to the question who the P Aryans were, one may fir& eay that they appetlr as a eecond etratum, 80 to say, covering a part of the area previouely occupied by a Q stratum, namely, from the North Sea ae far perhape aa Asia Minor, with an overflow into Britain, Spain, Italy and Greece. So their area of dispersion appears to have been included withiu the other area. Further, since the language of the P Aryans ia to be regarded tie a modified form of the older Aryan apeech, it may be aeked to what cause the modification is to be traced. One might of courae answer that all languagee are in a permanent etate of change, more or less rapid, eo long aa they are living; but this would hardly be a sufficient answer in the case of changes 80 definite ae those here in question. We have accordingly no alternative but to suppose, a8 already suggeeted, that the dialects of tbe P group aroee from the coming of speakers of the older Aryan dialects, namely, those of the Q group, in contact with a nowAryan race, which, conquered or otherwise powerfully influenced by them, adopted Aryan speech without having got rid of the non- Aryan characterietice of its inherited pronunciation. This eupposition of a very considerable ahaorption of non- Aryan elementa makee the P Aryans a mixed people talking what might be termed Neo-aryan. Thie view which derives counten- ance in this country from the fact that archmlogista find the

1 Book, i. 66. In the chapter following he surmises tbe djalecta of Creston and Placis to have been Pellwgion in hia OWP day. Phil. Traru. 1891-24. 9 130 CELTS AND OTHER ARYANS.-J. RHTS. round-barrow Brython .of the Bronze age to have been pret dominantly a broad-skulled man, though it is believed that the original Aryan waa long in this rmpect rather than broad or round. Perhaps one might venture to regard the lake- dwellings of Switzerland as the homes of the mixed population of the Neo-aryans. At any rate, we have an instance of the poseibilities of the Alpine region 80 late as the time of Csesar, when the Helvetii aet out from their country en maw to seek a home elwewhere ; and, but for the intervention of the legions of Rome, they would doubtlese have succeeded, aa many similar migrations from the Alps had probably done before. It is interesting to note the destination of the Helvetii: they set out for the territory of the Santones in tho West of Gaul: in other words, theirs was the march of Aryan8 of the P stock to conquer the territory of Aryans of‘ the other stock, or else of a people perhaps more correctly described aa not Aryan at all. Having endeavoured to show that the Aryans of the P etock emanated from a common centre, I have nothing to add except a word as to the wider classification suggested by Brugmann’s treatment of the consonants of the palatal and velar series.’ Compare, for example, the Gothic /toas ‘ who,’ with its Lithuanian and Sanskrit equivalent kns, which gives no iadication of its ever having been kyns or qys: similarly in the case of velar g and gh, the eastern group of languages ehow, according to Rrugmann, no trace of 9 as the mark of On the other hand, they agree in differing in their treatment of palatal k, g, and yh from the languages of the western group, namely, in that they reduce those mutes mostly into spiranta.g The Western GIroup consists of the following languages : Teutonic, Celtic, Italic and nellenic, and the eastern group of Letto-Slavic, Albanian, Armenian, Iranian and Sanskrit. So our P Ianguages take their place as a eub-

1 Grtmdrisr, vol. i. $§ 380-466. Ibid, $ 4 17. 3 Ilid, .j 380. P AND &.QROUPS. 131

I. LABIALIEINOLANOWAOES. 11. AWXBILATIHOLAXOUAOSS: i. Thoee characterized, wme time er other, by qp, gp, g&w, namely : LettoSlavic. 1. Teukmie. 2. a. Qoiddic. Albanian. b. Latin. e. Herodotein Qreek. Armenian. ii. Thoee with the p combinations reduced tothe labialsp, b, cp, namely: Iranian. 1. Brythonie. 2. Osw- Umbrian. Sanskrit. 3. Standard Qrnk. It is right to state that Brugmann, while distin- guishing .between the Aryan languages which labialize and those which do not labialize, hesitates to draw the concluaion that the Sryan parent speech had split up into two dialects. In his discussion of the consonanta, however, he is obliged to divide the whole family into the two groups, which have been termed respectively Weetern and Eastern, in spite of Albanian having somehow straggled out of the direction indicated by the geographical position of the other members of its group. It need scarcely be added that he deala with the Aryan languages and not with the ethnology of the nations by whom they have at any time been spoken. Since thei+e conjectures of mine were written an important article has been published in the Nineteenth Century by Prof. Huxley on ‘‘ The Aryan Question and Prehistoric Man ” ; and I may say that my view is decidedly favoured by his conclusions, though I cannot now enter into detaila, as my paper ie already longer than it was intended to be.

1 Ibid, 4 417, and note 1.