kimkat0369k Lectures on Welsh Philology. 1877. John Rhys (1840-1915).


● kimkat0001 Yr Hafan
● ● kimkat2001k Y Fynedfa Gymraeg
● ● ●
kimkat2194ke Cyfeirddalen yr Adran Ramadeg
● ● ● ●  kimkat0369k
Y tudalen hwn




(delwedd B60)

[AU rights reserved. ]



(delwedd B60)




(delwedd B60)




(delwedd B60)

TO ,



(delwedd B60)


The substance of these lectures was delivered at
Aberystwyth College in 1874, so that they were
intended to appeal, in the first instance, to Welsh
students of Celtic Philology ; but it is hoped that
they will also be found intelligible to other than
Welsh readers, and with a view to this the Welsh
instances have been rendered into Eaiglish through-
out. Since they were first delivered they have
been re-written almost entirely, and the author
could have desired to repeat the process ; but at
that rate publication would have been out of the
question, as his views are constantly undergoing
modification, which will surprise no one aware
how recently the systematic application of the
comparative method of study to the Celtic lan-
guages began. His excuse for publishing at all,



(delwedd B60)


under the circumstances, must be the fact that,
although the highest effort of one student may-
result only in giving him a glimpse- of half the
truth, even that may enable another to discover
the whole truth, and to secure for both a more
advanced point of view. The chances of his doing
this appear to outweigh the probability of the
crudeness of his theories leading others astray
who are not in the habit of trying to think for
themselves, persuaded as he is, that, if they do
not derive wrong ideas of Celtic questions from
these pages, there are plenty of others from which
they will. Besides, it would require a livelier
imagination, and more ingenuity than he could
boast of, to originate, with regard to the history
of the Celtic languages or nations, any theories
which could vie in absurdity and distorted vision
with many of those stUl obtaining among people
of the class mentioned.

The reader will have already surmised that the
Lectures do not form a harmonious whole : one
reason for this was the gradual comino- in of
more accurate knowledge about some of the
most important of our Early Inscriptions after
the -MS. had been in the printer's hands. The



(delwedd B60)


study of the former cannot fail to form an era
in Welsh Philology, and no inference warranted
by them could safely be overlooked. To a student
of Greek or Eoman epigraphy they might, it is
true, appear of little importance both in point of
meaning and of number, but meagre as they are,
to those who are desirous of understanding the
history of the Welsh language, they are simply
1 invaluable. The author has the satisfaction of
having, in the course of the last four summers,
inspected nearly all of those still preserved, to-
gether with others of a somewhat later period, of
which it was not thought necessary to submit g,
detailed account, seeing that they mostly belong
to the time of the Old Welsh Grlosses, and form
accordingly a part only, and that the less im-
portant one, of the available materials for the
study of Old Welsh.

As to the meaning attached here and else-
where in this volume to the terms Early, Old,
Mediceval, and Modern Welsh, the reader is re-
ferred to the beginning of the Fourth Lecture,
page 143. And by the frequently recurring
words, our Early Inscriptions, are briefly meant
the old inscriptions, not of Koman or English



(delwedd B60)


origin, whicli have been found in Wales, Devon-
shire, and Cornwall, together with one or two
in- Scotland that appear to belong to the same

Rhtl, January 1, 1877.



(delwedd B60)





Intbodcctort Sketch of Glottology — Grimm's Law-
Classification OP THE Celtic Languages . . 1

Welsh Consonants ... . . 36

Welsh Vowels ....'. .90

A Sketch of the Histokt of the Welsh Language . 140



Ogams and Ogmio Inscriptions 272


An Attempt to Eeconsteuct the History of the Ogmic

Alphabet 329



(delwedd B60)




A.— OuB Eablt Insckiptions .... 379

B. — Maccu, Mncoi, Maqvi, Maowt . . .415

C. — Some Welsh Names of Metals and Articles

MADE OP Metal 420

Additions and Coerections 433

Index . . ... . . 445



(delwedd B60)





" If we meet in difierent tongues with words which are clearly the
same word, notwithstanding differences of form and meaning which
they may exhibit, we cannot help concluding that they are common
representatives of a single original, once formed and adopted by a
single community, and that from this they have come down by the
ordinary and still subsisting processes of linguistic tradition, which
always and everywhere involve liability to alteration in outer shape
and inner content." — William Dwight Whits et.

If you glance at that part of the Old "World ex-
tending from the Ganges to the Shannon, and
consider the Babel of languages spoken within
that range, you will be able to form an idea of
the difficulty of satisfactorily classifying them.
However, that has been so far done, and with so
much success that the results are not likely to
be very gravely compromised by future investiga-
tions. Roughly speaking, we have within that
stretch of the Northern Hemisphere three great
families of speech, namely, the Aryan, the Semi-
tic, and the Turanian. The first, of which more
anon, comprises the idioms of the chief European




(delwedd B60)


nations, and of Hindoos, Persians, and Armenians.
The Semitic languages reckon among their num-
ber Arabic, Hebrew, Syriac, and kindred tongues.
As Turanian we are taught by some to treat
Turkish, Hungarian, Finnic, Lappish, Samoyedic,
and a number of other nearly related dialects
spoken in the Euss.ian Empire, to which may now
be added Accadian, one of the languages of the
Cuneiform Inscriptions of Ancient Assyria. This
covers a considerable portion of Asia and all
Europe, excepting the south-west of France and
the north of Spain, where Basque is still spoken,
a language whose place in the Turanian family
has not yet been made out. It is, however, cer-
tain that it is neither Aryan nor Semitic.

To return to the Aryan family with which we
are here more especially concerned, the analysis
of the languages, formerly or still spoken by the
leading nations of Hindoostan, Persia, and Europe,
has led to the conclusion, that they are, linguisti-
cally speaking, descended in common from a single
primeval tribe. So far all may be said to agree,
but not so when we come to the question as to
how and in what degrees the Aryan nations are
severally related one to another within the family
they make up. The older and still, perhaps, the
prevailing theory, which has found a doughty
champion in Dr. Fick of Gottingen, sets up a



(delwedd B60)


genealogical tree to the following effect : — The
original Aryan tribe broke up somewhere in the
neighbourhood of the Caspian Sea into two, where-
of the one, proceeding eastward, forced its way
ultimately into Hindoostan and Persia, while the
other made for Europe. Thus we have an Eastern
or Asiatic branch, and a Western or European
one. The former is represented by the Hindoos
and Persians, and the latter is supposed to have,
in the first instance, yielded a Northern and a
Southern division : the Northern Aryans of Europe
comprise the Teutons and the Litu-Slaves. The
Teutons include the Aryan nations of Scandinavia
and Iceland, the High Germans, and the Low
Germans, among whom our nearest neighbours,
the English, are reckoned. The Litu-Slaves fall
into two groups, whereof one includes Lithuanians
and Letts on the Baltic in a country divided be-
tween Prussia and Russia ; not to mention the
Old Prussians or Borussi, who inhabited parts of
Prussia now completely Germanised, and gave their
name to Prussia itself, and to Berlin and other
towns, where their memory is now a mere matter
of history. The other group comprises the ruling
race in Russia, Poles, Servians, Bohemians, Wends,
and other nearly related races located within
the areas of the Russian, Ottoman, Austrian, and
German empires, and forming the disjecta membra



(delwedd B60)


of a Slavonic world not easy to define without t
aid of a good linguistic map of Europe. T
other or Southern division of the European Arya
comprises — first, the Greeks and allied raci
forming a whole with its centre of gravity son
where between the Adriatic and the Hellespon
secondly, the Italians, who speak a variety
Eomance dialects, preceded in Ancient Italy by
less a variety, including, among the most ii
portant, Latin, Oscan, and Umbrian — the affinit
of Etruscan are still, owing to the diŁ3culty of i
terpreting its remains, subjudice: it willprobal
turn out to be non-Aryan. And, thirdly, t
Celts, called by the Romans Galli, by the Gree
KsKtoi and TaKdrai, and by themselves, or, rathi
by those of them who inhabited Gaul or Ancie
France, according to Caesar's account, Celtse, as
whom it may be said that some three hundr
years before the Christian era, they occupied t
British Isles, Gaul, Switzerland, a part of Spai
South Germany, and North Italy : not long afl
some of them passed into Asia Minor and ga
their name to the province of Galatia.

The advocates of this theory are in some troul
as to how to deal with these three groups ; t
difficulty being, that Latin and the Celtic la
guages are so similar in many important respec
that they are not to be severed, while, on the oth



(delwedd B60)


hand, Latin and Greek are still more closely allied.
The consequence is, that some subdivide the
Southern division into an Italo- Celtic and a Hel-
lenic group, while others prefer to suppose a Celtic
and a Greco-Italic group. This is one of the dif-
ficulties of the genealogical theory ; but there are
a good many more under which it labours, and
which have been formulated by Johannes Schmidt
in the first part of his book entitled Die Verwant-
schaftsverhdltnisse der indogermanischen Sprachen
(Weimar, 1872), in which he propounds his own
views. The latter I could not better describe than
by rendering, as literally as I can, his own words :
a paragraph beginning on page 28 runs thus: —
" The figure also of an inclined plane dipping in
an unbroken straight line from Sanskrit to Celtic
appears to me not inappropriate. As to linguistic
boundaries within this range, originally there were
none : two dialects A and X taken at any distance
you please apart in it were connected with one
another by the continuous varieties B, C, D, &c.
The appearance of linguistic boundaries, or, to
abide by our figure, the transformation of the in-
clined plane into a flight of steps, I look at in this
way : — one family or one stem speaking the variety
F, for instance, gained, for reasons political, reli-
gious, social or other, the upper hand over its
immediate neighbourhood. Thereby the nearest-



(delwedd B60)


lying varieties of speech, G, H, I, K, in the one
direction, and E, D, 0, in the other, were sup-
pressed by F and replaced by it. After this had
happened F bordered immediately on the one side
on B, on the other immediately on L : the varie-
ties connecting both were on the one side raised
and on the other sunk to the level of F. Thus a
sharp linguistic boundary had been drawn between
F and B on the one hand, and between F and L
on the other, a step taking the place of the in-
clined plane ; and surely this kind of thing has
come to pass often enough in historical times. I
will mention only the influence of Attic as it grew
stronger and stronger, and gradually drove the
dialects quite out of the field of Greek literature,
the language of the city of Eome suppressing the
other Italian dialects one and all, and Modern
High German destined, and that perhaps at no
very "distant a date, to bring about the like extir-
pation of the German dialects."

These languages, whether, in the task of classi-
fying them, one follows the lead of Fick or of
Schmidt, are known collectively by various names,
such as Japhetic, Indo-European, Indo- Germanic,
Indo-Celtic, Aryo-European, and simply Aryan,
none of which are free from objections, but Aryan
recommends itself by its brevity. It is, however
to be remembered, that it is usually confined to



(delwedd B60)


the Asiatic brancli, the Aryans of India and Iran,
by Continental writers, who, in case they are
Germans, call the entire family Indo- Germanic,
while a natural antithesis has suggested to the
French mind the compound Indo- Celtic. Aryo-
European, though also a new-fangled term, is more
logical than Indo-European, which is still very
commonly used here and in France : Japhetic
seems to be out of favour and old-fashioned^
though quite as good a term as Semitic, which
continues to be applied to another great family.

To pass from this question of names to another
and a more important one, it may be asked how it
is known that the Aryan languages are of one and
the same origin. In answer it may briefly be
said, that one of the readiest ways of satisfying
one's self on this point is to compare the voca-
bularies of the languages in question, especially
•the more permanent portions of them, such as the
pronouns, the numerals, and the terms expressive
of the nearer removes of blood-relationship. Thus
nobody can fail to see to what conclusion the simi-
larity between the following words must point : —
Welsh mi, Irish md, Latin me, Greek fie, Eng.
me, Lithuanian manS*, Old Bulgarian (so the Sla-
vonic language of which we have the earliest speci-
mens is called) »ze", Sansktit m&m, Zend m&m ;
Welsh dau, Irish da, Latin duo, Greek Ivo, Eng.



(delwedd B60)


two, Lith. dii, 0. Bulg. dva, Sansk. dm, Zend
dva; "Welsh brawd, Ir. brdthair, 'L2X. fr&ter, Greek
<f)pdTrjp ' a clansman,' Eng. brother, Lith. broterdlis,
•0. Bulg. hratru, Sansk. bhrdtar, Zend brdtar.
Suffice it to say, that, if you chose to carry this
simple inspection far enough, you would probahly
find they instances at your command so many and
such as to preclude the possibility of their simi-
larity to one another being the mere result of
accident or of borrowing. Should you still hesi-
tate to ascribe their similarity to a common origin
of the languages they respectively belong to, there
remain the irresistible arguments -which the gram-
mar of the latter never fail to supply. That is,
in a few words, the kind of reasoning on which
comparative philology, or, as it has been more
concisely called, glottology, may be said to be
mainly founded ; at any rate, so far as concerns
the leading families of human speech.

In passing, one cannot abstain from calling at-
tention to the historical value and importance of
the method of glottology already mentioned. A
few specimens will serve to show how it lifts the
veil of darkness which conceals from our ken the
antiquity of the race. Thus from Welsh ych ' an
ox,' plural ychen, Breton oc'hen, Eng. ox, oxen,
Sansk. ukshan, 'a bull,' it is concluded that the
primeval Aryans had a word uksan meaning an



(delwedd B60)


OX or bull ; and from Welsh hu, bum, bumch, ' a
cow,' Irisli bo, Lat. bos, Greek /3oi/?, Eng. cow,
Sanst. ffo, that they had a word ffvau meaning a
cow or an ox : hence it is evident they were*
familiar with horned cattle. In the same way it
could be shown that they had horses, sheep, goats,
swine, and dogs. They lived not in tents, but
in some kind of houses with doors to them [Welsh
drrvs, Ir. dorus, Greek 6vpa, Eng. door, Sansk.
dvdrd], and they knew how to kindle [Welsh
enni/n ' to kindle a fire,' Sansk. indk the same,
indkana ' firewood, fuel '] fires in them. Those
fires served to make their pots or cauldrons boil
[Welsh pair ' a cauldron,' Med. Ir. coire, Sansk.
caru] : in them they cooked and stirred about some
kind of broth or porridge [Welsh umd ' porridge,'
Breton ioi, 0. Irish itk " puis," Lat. jus ' broth,
soup,' Greek fm/io? ' soup,' Lettish jdut ' to stir
meal about in water,' 0. Bulg. jucha ' soup,'
Sansk. yus, yusha, 'broth, soup']. What kind
of meal entered, into the composition of this /leKa^
^a)IJ,o<! is not known, as the evidence bearing on
their skill in agriculture is very scanty. But that
they had some kind of corn is proved by the
. equation of the Welsh word haidd ' barley ' with
Sansk. sasya, Zend hakya 'corn, a field-crop.'*

When this was suggested to Mr. Whitley Stokea, he kindly
called my attention to the following passage in Pliny xviii. 40 : —
"Secale Taurini sub Alpibus asiam vocant"— he proposes to read



(delwedd B60)


They did not go naked, but wore clothing [Welsh
gwisc, Lat. vestis, Sansk. vastra\, made probably
of wool [yf elsh. gwlan, Lith. vilna, 0. Bulg. vluna,
Eng. wool^ Sansk. urna\. All this tends to show-
that they lived in the North Temperate Zone, that
is, as it is supposed, in Western Asia, far away
' probably from the first man's abode, sometimes
assigned by anthropology, in its attempt to grapple
with the difficulty as to how Australians, Coolies,
Papuans, and Negroes reached their respective
homes, to a continent which it undertakes to pro-
ject as once extending from Africa eastward by
Madagascar and Ceylon as far as Celebes. But
although we read in the Book of Genesis how
Adam was driven out of Paradise with its four
mysterious rivers, they are, perhaps, a little san-
guine who expect that deep-sea dredging in the
Indian Ocean may one day be the means of
bringing to light a twig or two of the tree of
knowledge. Now that our inquiry is overtaken
on a by-path, it is liable to be waylaid by the
evolutionist and stopped by the theologian; the
former wishing to know how far our Aryan fore-
father had risen above the ape, and the latter how
far he had gone from original righteousness. The

lasiam. Further, in his Remarks on the Celtic Additioni to CfuHius'
Greek Etymology, &o. (Calcutta, 1875), p. 43, he points out in the
two first letters of the Irish word eorna 'barley,' the rule-right
Irish representative of Greek feid, Lith. javai, Sanskrit ywva.



(delwedd B60)


answer has been partly given already : the following
remarks may he added: — Looked at from an intel-
lectual point of view, we do not know much about
him beyond the facts, that he could count as far as
one hundred [Welsh cant, 0. Ir. cdt, Lat. centum,
Eng. hundred, Sansk. qata\, that to him to know
was to see [Welsh {yn dy) wydd (' in thy) sight,'
gwyddost ' thou knowest,' Lat. video ' I see,' Greek
olBa ' I know,' elhov ' I saw,' Sansk. vedmi, veda, ' I
know '], and that he knew how to stretch and touch
a number of strings so as to elicit from them music
to cheer his leisure hours or to enliven his festivi-
ties [Welsh tant 'a rope, a string, a musical string,' )J /Q/^ ^
plural tannau ' the harp,' 0. Ir. t^t {gl. fidis), Sansk. —
tanti, tantu, ' a string, a chord,' tata ' a stringed in-
strument ; ' Greek tovo'; ' a rope, a cord, a strain,
tension, a note, a tone,' rao-i? * a stretching, a
raising the pitch in music '].

Socially he seems to have been the master of
his house on a footing of equality with his wife,
who was mistress of the same and not a slave.
His children not only addressed him as father, but
they also called him more familiarly tata [0. Welsh
tat. Mod. Welsh tad, our only word for father,
Lat. tata, a fond word for father, Greek rdra. Terra,
0. Bulg. tata, Sansk. tata, tdta']. His vocabulary
appears to have been very copious as regards the
various ramifications of the family, whence it is



(delwedd B60)




inferred that each individual had his standing in
it well defined, a state of things highly natural in
a patriarchal system of government. His ideas of
religion and morals can only be guessed, and how
many gods he had it is impossible to say. It is,
however, certain that he worshipped one above all
others, if others he had, and that he spoke of him
in terms expressive at once of the light of day and
of the wide expanse of the sky, which looked down
upon him wherever he roamed [0. Welsh diu
' God,' Med. Welsh diu ' day,' Mod. Welsh Duw
' God,' dytv 'day,' ke-ddyw ' to-daj,' Ir. dia ' God,'
in-diu ' to-day,' Lat. Diovis, Jovis, deus, divus,
sub divo = sub Jove ' beneath the open sky,'
Greek Zew, genitive Alo^, Stos ' heavenly,' evStos
' at midday,' Sansk. div, dyu, ' the sky, day, bright-
ness ']. This may have been merely his way of
saying that his great Heaven-father [ = Lat. Dies-
piter, Joupiter, Jupiter, Juppiter = Greek Zev
Trdrep = Vedic Sansk. Dyaushpitar\ was the god
of light, and that he was present everywhere.
Whether he worshipped light or not, as such, in
the performance of his religious rites he seems to
have been in the habit of standing with his face
turned to the rising of the sun and his right hand
to the south [Welsh dehau 'right (hand), south,'
Deheu-dir 'the south land, i.e., South Wales,'
0. Ir. dess, Mod. Ir. deas ' right, south ; ' the



(delwedd B60)


Teutonic instances are Teisterhant and Texel, in
which the first syllable is supposed to mean south
(see the Eevue Celtique, ii. p. 1 73) ; Sansk. dak-
shina ' right, south,' dakshind {dig) a southern
country supposed to be the Deccan]. . A ( '- -'/
How he stood with his god or gods it is im-
possible to say, but he seems to have been no entire
stranger to his own shortcomings, and the con-
sciousness of some kind of sin or guilt, as proved
by Welsh euog 'guilty' (for other instances of
Welsh eu = ag see the Rev. Celt., ii. p. 193), Greek
ayos ' pollution, guilt, a curse,' Sansk. dgas
' offence, mistake, transgression,' words which
bring into a strange rapport with one another the
disciples of Buddha in the far East, the followers
of Calvin in Wales, and those subtle Greeks of
old, in whose history, religious and political, the
ayos played a conspicuous part. The natural corol-
lary to this is the inference that the religion of
out Aryan ancestors must have had its ascetic side,
and enjoined on them some kind of penance and
self-mortification, as suggested by the following
words : — Welsh crefydd ' religion,' meaning for-
merly religion from the point of view of an ascetic,
whence crefyddrcr in the Middle Ages meant a
religieux rather than a religious man in the ordi-
nary Protestant sense : Irish crdibdech ' pious ' (in
the Book of Armagh), craibhtheach ' religious,



(delwedd B60)


devout,' craibhdhigh " people who mortify the flesh "
(O'Reilly), Sansk. ^ram, cr&mya ' to become tired,
to labour in vain, to chastise one's self,' grdnta
(for crdm-ta) ' fatigue, pains, chastisement, the
result of religious effort,' gramana ' one who chas-
tises himself, an ascetic, a beggar-friar, a Buddhist,'
gramand ' a beggar-nun,' agramana ' an anascetic'
So, after wandering about in the mists of anti-
quity, we unexpectedly find ourselves near a point
conspicuous in the religious landscape of our
own day.

When we set out on this digression we were
considering the phonetic similarity of cognate
words belonging to different languages, but in the
course of it instances were intentionally brought
together, which may, on the other hand, have
forced their differences into relief. It will, how-
ever be some consolation to find that the majority
of those differences follo\7 fixed rules. Thus, to
recall the Welsh word pair and the Irish coire,
the same p-c variation occurs in other cases, such
as Mod, Welsh pen ' head,' pren ' a tree,' pwy
' who,' and 0. Welsh map ' a son,' which are in
Irish ceann, crann, cia, and mac, respectively.
Similarly in equating Welsh cant with Eng.
hund-ied we assumed Welsh c to be represented in
the Teutonic languages by h ; and that is found to
hold true in other instances : take Welsh caj'-a.el,



(delwedd B60)


Eng. ham, "Welsh cos ' hateful,' Eng. hate, Welsh
ci ' a dog,' pi. cwn, Eng. hound, "Welsh coed ' a
wood,' Eng. heath, "Welsh coll, Eng. hazel, "Welsh
craidd ' centre,' Eng. heart. Now it is one of the
characteristics of the Teutonic languages that they
deviate as regards the consonants in a consistent
and well-defined manner from the other Aryan
languages, and it is to the students of the former
that we owe the discovery of the rules alluded to,
or at any rate the more important of them. Hence
they are commonly called, after the scholars who
made them out, Grimm's Law, and sometimes
Eask's. By means of that law, and the other data
afforded by a careful comparison of all the more
important Aryan languages, some glottologists
think it possible approximately to infer both the
vocabulary and the inflections of that older lan-
guage whence they have all sprung. An idea may
be formed of the amount of work attempted in this
direction from the fact that the second edition
of Schleicher's Compendium of the Comparative
Grammar of the Indo- GermMnic Languages (Wei-
mar, 1866) makes up 856 pages octavo, and that
the second edition of Fick's Comparative Dic-
tionary of the same (Gottingen, 1870) covers
no less than 1085 pages octavo, while the third
edition, now publishing, is likely to occupy a good
deal more than double that number.



(delwedd B60)


Perhaps I could not do better than close these
preliminary remarks with a rough summary of
consonants etymologically equivalent in some
of the leading Aryan languages. No attempt is
made to make the table exhaustive by crowding
into it exceptional details, unless they happen to
be of special interest to the student of Welsh.
However, it will be found sufficiently exact to en-
able you with ease to bring to book many of the
fanciful etymologies which are ever floating about
in the atmosphere of Celtic philology until they
are caught by some reckless writer of the fantastic
school of history, who dearly loves wild specu-
lations on the past of some one of the Celtic
nations ; for a false etymology can seldom be said
to be insured against speedy oblivion until the
Muse of History has taken it by the hand and
assigned it a sphere of usefulness.

Now that you have a general idea of the way
the student of comparative philology goes to work,
and the position which the Celtic languages oc-
cupy in the. family to which they belong, your at-
tention must be called a little more in detail to
them. It has already been hinted that they offer
more important points of similarity to Latin and
its sister dialects of Ancient Italy than to any
other Aryan group of languages whatsoever^ —
herein Fick and Schmidt would agree ; but next,

o S^ a> z o

s * a *tŁ

,.- •-» SB m

JS 2 o S a



(delwedd B60)



. fc* > J3 r^ ri3 ja bo . .



— to r-" '^ =r '-.


^^ bo _" 1 f

oo-^ : bo-o T3 -o bo-o TS ^ c a a tT :*_-^ar


<^ ***" bO « '"'

o eu-w " bo-o 73 ^ tio-J3 Td ^ c a E tT-r-, > m

poB uraao

bo rt —



I luapoH


j3 S:3>hJ=; ct<^ q. to bcTj ^ = a S !.">>&»"'



M^ bo r! ""^

ffi 4^ ^ fU'N bc^ r^ >^q bo^ ^ a a a ^ 'r^ > m


» jf * !% H bO-O J5 N bO-O J3 1 a" a -".rri !> aT


OiJiT-ia P* N !Łt3 ,Ł3 N bO'^ r^^ id B U >i > ^


O.J4-B o,^ bo-e ,Q Ja i^Cl,:2.'a a fi tT t>. t> »

w at< fi< o o p M o p o m l2i ^ S p3 1>< ^ OT

» a S-u-s

ń3 Ł•"'0 s

"s-i SB-"

•^ m a ;" S 5j-

a S.2* Stj

Ł«§!" = «

» s K-^ ^

a » p a o S
,a^^ Bsg

c -.2 Sap.

r-fe a •5 p









(delwedd B60)


perhaps, to the Italian group they loost nearly
resemble English and the other members of the
Teutonic group. This fact, which is gradually be-
coming more evident as Celtic glottology progresses,
is fully taken into account by the dialectic theory,
as coinciding with the geographical position oc-
cupied at the dawn of history by the Celts between
Italians and Teutons ; whereas the genealogical
tree would lead one to expect to find them resem-
bling, in point of language, the Slavonians quite
as much as the Teutons, which is certainly far
from being the ease. It is also to be noticed, that
it is owing to the encroachment of languages de-
riving their origin from Italy and Germany, that
the vast Celtic world of antiquity has been, as far.
as regards language, reduced to its present narrow
dimensions, that is to say, the fag-ends of France
and the British Isles. This is, however, an aspect of
our history which no one could expect us to dwell
upon with feelings of pleasure and satisfaction : as
we believe the Celts never to have been cowards,
we turn away fain to think that the words which
the poet makes Hector apply to individuals hold
equally true of races : —

Moi/jax Soinvd <p7]fu vcipvyixivov l/i/ieiiai i.vSpSiv,
Qi KaKbv, oiSi fiev i<r$\ip, iir^v t4 vpHra yivrirai..

The Celtic languages still spoken are Welsh,
Breton, Gaelic in Ireland and the Highlands of



(delwedd B60)


Scotland, and Manx : among the dead ones are Old
Cornisli, Pictish, and Gaulish. Of these, Cornish,
which ceased to be spoken only in the latter part
of the last century, has left us a considerable
amount of literature, while the Pictish words
extant may be counted on one's fingers : the
old Gauls have left behind them a number of
monuments, from which, together with other
sources, a fair number of their names and a few
other specimens of their vocabulary have been col-
lected; enough in fact to enable one to assign
them their proper place in the Celtic family.
Now as to the Celts of the British Isles and Brit-
tany, all are agreed that they divide themselves
naturally into two branches, the one Kymric and
the other Goidelic. To the latter belong the
Irish and the Gaels of Scotland, together with the
Manx ; to the former the Welsh, the Cornish, and
the Bretons, not to mention that the Picts, Mr.
Skene notwithstanding, were probably Kymric
rather than Goidelic. Then as to the Ancient
Gauls, it has been usual to range them with the
Kymric nations, so that you will find the entire
Celtic family commonly spoken of as consisting
of Goidelic nations on the one hand, and Gallo-
British ones on the other.

There are, however, good reasons for regarding
this classification as resting on a bad foundation,



(delwedd B60)


namely, a phonological argument which will not
bear examination. It is this : The Welsh and the
Gauls belong to the same branch of the Celtic
family, because their languages agree in replacing
Aryan qv by j», while Irish uses c : thus the 0.
Irish word ioxfour was cethir, while our word is
pedrvar, formerly petguar, and the Gauls called a
kind of carriage in use among them petorritum, a
form which no doubt involves their word fovfour.
The corresponding Latin, it is needless to add,
yjSLS quatuor, and the Aryan original was probably
qvatvar. Now a glance at the equivalents of
Aryan qv in the table will serve to show that this
kind of reasoning, if it proves anything, proves
rather too much. For why, it may be asked,
should the Welsh not be asserted also to be par-
ticularly near relations of those Italians, for in-
stance, who said petur for quatuor, of the Greeks
who called the same numeral irivvpe';, and of the
Modern Koumanians who have modified the Latin
words aqua and equa into ape and eape respectively ?
That would of course be absurd, and it is evidently
dangerous to rest a theory of history or ethnology
on such a basis. Nor is this all : the p coinci-
dence between Welsh and Gaulish should imply
something like an identity of date ; that is, both
languages ought to have had p for qv in use at
the same time, so as to allow one to infer that qv



(delwedd B60)


had become jo at a time when they were as yet one
language. This would be another twig of the
genealogical tree, and a contradiction of the facts
of the case. The Gauls had replaced qv by p at
some date anterior to the time of Caesar, whereas
our ancestors do not seem to have done so much
before the 6th century. You will have already
learned from the table, that Aryan p had disap-
peared from the Celtic languages : so, previous to
the change by the "Welsh of qv into /», the latter
sound must have been unknown to them. Accord-
ingly we find that the Ogam alphabet made no
provision for it, and that, when our ancestors be-
gan to borrow Koman names with j», they had to
invent symbols for it : more strictly speaking, they
seem to have extemporised them, for in the only
two instances extant they are different ones.
The former are pvnpkivs, accompanied by Pope-
in Ogam, on the Cynffig stone, and tvepilli,
the genitive of the Welsh form of Turpilius, on
the Glan Usk Park stone near Crickhowel : the
Ogam is not easy to read, but Turpil- is certain.*
, The other names with p, Pascent-, Paternini,

* The Ogmic symbol for p in Turpil- is of the form of x placed
on the right of the edge. The same symbol placed on the edge has
lately been proved by Dr. S. Ferguson to occur fqr^ in an Irish in-
scription reading : Sroinienas poi netattrenalugoa, which Mr. Stokes
would treat as Broinioonas poi netat Trenaliigos, and render, literally,
(Lapis) Broinidnis (qui) fuit propugnatorum TrenalugHs : see the
Proceedings of the RoyaJ, Irish Academy, vol. i. ser. ii. pp. 292-297,



(delwedd B60)


Paulinus, Potenina, which occur on our older class
of monuments, are unfortunately not given in Ogam.
The earliest native name in point is pobivs in
debased Eoman capitals hardly older than the
6th century, and the earliest instance in manu-
script is the related name borne by the king of
Dyfed in the time of Gildas, who uses it in the
vocative case as Vortipori ; Gildas wrote about
the middle of the 6th century. Kymric names
with qv are more numerous, and, probably,
earlier : in debased capitals we have for in-
stance maqv[eragi], maqvirini, QVENATAvcr and
QVENTENDANi, of which the last- mentioned is
A highly interesting instance : it seems to be
a derivative from Qvenvend-, which in Modern
Welsh is peuTvyn, ' white-headed,' and as a pro-
per name T Penwyn, ' The Whitehead.' In Irish
this is Cennjinn, whence is formed Cennfin-
nan, which is, letter for letter, our Qvenvendan-i,
and has its parallel in the Irish name Cenndubhan,
similarly formed from dubh, ' black : ' we may com-
pare in Welsh Carnwennan, Arthur's knife, from
carnwen, ' white-hilted.' Nor is this all, for Pen-
non and Cennjinn find their Gaulish representative, •

where also another stone ia mentioned by the Bishop of Limerick
as reading : Carbi poi macui Ldbradi — Mr. Stokes would render it
(Lapis) Corbi (qui)fuit gentis Lahradii. These excellent suggestions
of Mr. Stokes I have taken the liberty of publishing from his letters
to me last May.



(delwedd B60)


to which my attention was lately called, on a
silver coin in De Saulcy's collection (^Reo. Celt.,
i. 297) in the form nENNOoviNAoc, i.e., Pen-
nowindos. In Ogam we frequently have maqvi
the genitive of the word for son, and an inscrip-
tion from Devon reads Swaqqvuci maqvi Qvici,
which deserves a word of explanation. Swaqqvuc-i
is probably a derivative from srvaqqv-, which must
be the prototype of Mod. Welsh chwaff, used in S.
Wales in the form hwaff or waff, and meaning
' quick, quickly ; ' and as to Qvic-i, the same name
occurs in Irisb Ogam written Qweci, for that is
how I would read ' ' ' ' ' , , , 1 1 1 m^hh-hh. , As a rule,
however, our qv is so written also in Irish Ogam,
as in maqvi, which occurs scores of times on
Irish monuments written maqvi, maqqvi, moqvi,
with a single Ogam, j_l±jjl, for qv, or doubled for
what I transliterate qqv. But in the earliest
specimens of Irish and Welsh found in manuscript
Irish qv had been simplified into cc or c, and
Welsh qv made into p, so that the word for son
became mace or mac on the other side of St.
George's Channel, and map, now mab, on this —
the G-aulish cognate is supposed to be the simple
form implied by the Gaulish derivative Mapilus
(Kuhn's Beitraege, v. p. 364). To talk of the
Welsh changing c into p, it is almost needless to
remark, is the result of ignorance of the laws of



(delwedd B60)


phonology : where Irish c and Welsh p are equi-
valent, they both represent an earlier qv which, it
is interesting to notice, the Irish retained intact
till after the time when the Welsh began to change
it into p. Thus Irish hagiology speaks of a St.
Ciaran, whose name it also preserves in what is
evidently a much older form, Queranus. He is
supposed to have lived from 516 to 549, and to
have been the first abbot of Clonmacnoise. There
was, however, an earlier Irish saint of the same
name who was born in St. Patrick's time, and is
supposed to have died in the year 600. Fortu-
nately for our inquiry, he came over into this
country, and his name became modified into Fir- '
anus or Piran ; and a church in Cornwall still
bears his name, Piran in the Sands, Piran in
Sabulis or Peranzabuloe. Thus it would seem
that the Welsh were in the habit of changing qv
into p about the end of the 5th or the beginning
of the 6th century, while the Irish retained it
intact so late at least as the middle of the latter
century : so the Grallo-British theory can derive no
support from this quarter.

Were one inclined to use an argument like the
one which has just been condemned, one might
urge that Irish and Gaulish having initial s where
Welsh has h, makes for a Gallo-Goidelic unity.
This would of course be idle, as it is certain that



(delwedd B60)

ńLECTURE I. ' 25

onr ancestors changed s into k subsequently to
their borrowing from the Romans the word sex-
tarius, which they had made, before the end of
the 9th century, into hestaur, written later
hestawr, whence hestoraid, colloquially curtailed
in some parts of N. Wales into stored, a measure
of capacity of about two bushels. It need hardly
be added, that our early monuments never show
an initial k, but always s; but the process of
changing s into h in Welsh would seem to have
become obsolete before the middle of the 6th
century, if we may depend on the tradition which
refers the church of Llansannan in Denbighshire
to the Irish saint, Senanus, who is supposed to
have spent a part of his life in this country, and
to have died in the year 544 : this is, however,
not a very conclusive argument, as some native
words do not change s into h: take for instance
the numeral saith, ' seven,' and there may have
been reasons unknown to us why a foreign name
should not follow the rule obtaining in Welsh :
the double n also in -Llansannan creates a diffi-

Having severed the supposed Gallo-British ties
of special kinship, we are at liberty to re-classify
the entire family into two branches, whereof the
one embraces the Celts of the Continent, and the
other those of the Islands. This, however, does



(delwedd B60)


not in any wise interfere with the probability of
the Continental Celts having invaded this island,
and taken possession of extensive tracts of it long
after they and the Insular Celts had differentiated
themselves in point of language and history. In
fact, it is certain that parts of the South of
England had been thus occupied by invaders from
the Continent, among whom there were probably
Celts, if indeed they were not wholly Celts, be-
fore Julius Caesar landed' here. And if the com-
mon reading of a passage in Ptolemy's Geography
is to be depended upon, which mentions a people
called Uaplaoi, living in a town called JJerovapia,
near the Humber, one can hardly avoid drawing
the conclusion that the Gaulish Parisii had sent a
colony here. This is by no means impossible,
considering the position of the Uapicrot near the
Humber, and the possibility that the Parisii,
whose chief town, Lutetia, stood on an island in
the Seine, on a site still occupied by Paris, had
ships at one time at their command. And here
the following points, which I copy from Smith's
Dictionary/ of Greek and Roman Geography, are per-
haps not all irrelevant. It seems that the Romans
had a fleet at Paris ; a ship appears in the arms of
the city ; an inscription was dug up at Notre Dame
in the last century, reading Nautce Parisiaci ;
and the Senones, the neighbours, and probably



(delwedd B60)


the allies, of tlie Parisii, possessed ships; for
Csesar (vii. 58) states that Labienus seized about
fifty of them at once at Melodunum, higher np
the river. Whatever the noun implied by UeTovap'ia
may have been, the word is probably to be equated
with Mod. Welsh pedwaredd, ' quarta.' Brittany,
it is needless to say, is a kind of a counter-colony,
the Bretons being the descendants of countrymen
of ours who passed thither about the 5th century,
and not the direct representatives of the Ancient
Gauls, as is proved by their traditions and lan-
guage, which is a Kymric dialect easily learned
by a Welshman. I gather, however, that a lead-
ing French Celtist, M. H. d'Arbois de Jubain-
ville, takes for granted the Gaulish descent of
Breton ; but so far I am not aware that he has
made it the subject of special discussion. In the
same light as the British colony in Armorica, one
might also regard the settlement in Scotland of
Gaels from Ireland.

It is clear that the old classification, if it is to
stand, must be placed on a firmer foundation,
which, I am persuaded, is not likely to be dis-
covered. Nevertheless, it is impossible to prove
to a certainty that the one here proposed in its
stead is the correct one. At first sight it might
appear to be demonstrated as soon as certain
traits have been pointed out in which Welsh and



(delwedd B60)


Irisli agree with one another and differ from
Gaulish. But it is not so, as two languages may
take the same path independently of one another :
such points of similarity — and such there are in
spite of the scarcity of the Gaulish data — can
only yield a greater or less degree of presumption
in favour of the closer kinship of Welsh and
Irish. This is, however, a sufficient reason for
briefly mentioning a few of them.

A single s flanked by vowels is lost in Welsh
and Irish, but retained in Gaulish ; as in the
Irish word ffa, genitive ffai or ffooi, ' a spear, a
javelin.' Its Gaulish equivalent is gaesum, men-
tioned for instance in Virgil's description of the
followers of Brennus : —

duo quisque Alpina coruscant

Oaesa manu, tcutis protecti corpora longii.

In classical Latin the stems of nouns in the
second declension end in u in the nominative, as
in equuSf^lius, donum, but Old Latin equos,filios,
donom, on a level with the Greek uL?, 6e6<;, and the
like. The corresponding vowel in Sanskrit is a,
as in Qivas, ' the god Siva,' Mntas, ' car us,' kdn-
tam, ' carum : ' it is a also in Zend and written
Lithuanian, and it is generally considered to be
older than the u and o of Latin and Greek. The
inscriptions of Ancient Gaul show Gaulish to have
been in this respect on the classical level : witness



(delwedd B60)


the following forms : — Andecamulos, Cernunnos,
Contextos, Crispos, Doiros, Dontaurios, Iccavos,
OviXKoveo^, Seyo/jMpo^, Seviros, tarvos, Tarbelinos,
Ulcos; Brivatiom, canecosedlon, cantalon, celic-
non, Dontaurion, lubron, veixrjrov, Kamedon. The
evidence of the leading elements in compounds is
to the same effect : Danno-tali, OvivBo-fiayo^,
Samo-talus, Sego-mari, Yerno-dubrum. But on
the whole the early inscriptions of Wales and Ire-
land make for a — unfortunately we are nearly
confined to the leading elements in compounds :
Welsh — Cata-manus, Corba-lengi, Cuna-cenni,
Ena-barri, Netta-sagru, Trena-catus ; Irish — Ana-
dovinias, Cata-bar, Cuna-cena, Cuna-gussos, Eva-
cattos, Netta-lami: to this I would add an in-
scription from Ballintaggart reading Tria maqva
Mailagni, probably for Tria{m) Maqvam Mailagni
= TpiMv vlwv Mailagni. It is right to add, that in
the period to which our earlier Welsh monuments
are to be referred the vowel ending the leading
elements in compounds had got to be indistinctly
pronounced, a preparation to its entire elision in
later Welsh generally. In our bilingual inscrip-
tions a is used in Ogam, but advantage is some-
times taken of the obscure sound of the vowel to
write it o in the Latin version, or even e, which
tends to make the names look a little more like
Latin. Thus we find together Cunatami and Cuno-



(delwedd B60)


tami, Cunacenniwi and Cunocenni, Trenagiisu and
Trenegussi: also in two distinct inscriptions in
Roman capitals Senomagli and Senemagli. But on
the whole the weight of evidence is in favour of
the claims of a. Welsh and Irish inscriptions con-
tain derived forms ending in the genitive in gni :
Welsh — Maglagni, Ulcagnus, Corbagni, Curcagni;
Irish — Artagni, Corbagni, Dalligni, Mailagni,
Talagni, Ulccagni. In Gaulish names the same
suffix is cnos, cnon, genitive cni, as in celicnon,
Oppianicnos, ToovTicratKvo';, Druticnos, Druticni.
On a bilingual stone Druticnos is rendered Druti
Jilius, but the inference, that Gaulish had a word
cnos meaning son, is as warrantable as if, from
Ile\o'irihri<; = IIiko'Ko<; vlo<;, one concluded that Greek
had a word t8r}<; = vto?.

When the Celts first took possession of these
islands, it is highly probable that the patriarchal
system of government obtained among them, and
that it continued to flourish as a well-defined system
of tribes or clans, such as we find in later times
in Ireland and Scotland, during the long interval
between their coming here and their separation
into Kymric and Goidelic nations. And it is per-
haps to this prehistoric period of Goidelo-Kymric
unity that one is to refer the composition of most
of the personal names containing the word teyrn,
'a king, a monarch,' 0. Irish tigerne; now tighearna,



(delwedd B60)


' a lord :' in our early inscriptions we have tigirn-i
and tegern-o-. The etymon is the Celtic word for
house, which, in 0. Welsh, was tig, now ty, 0. Ir.
teg, now teach, genitive tige, now tighe : so the
word teyrn is perhaps an adjectival formation which
may originally have meant connected with or relat-
ing to the house, but in what special sense it is now
impossible to say. Its use was not confined to
the Insular Celts, for Gregory of Tours mentions a
Tigernum " castrum urbis Arvernse, Tigernense
castellum" — I am quoting from Gliick's Keltischen
Namen (Munich, 1857), p. 180 : in Auvergne this
is now Tiern, and the name is known to all in its
form of Thiers. Now the Celts of the British
Isles seem to have applied the adjective to the
householder or the head of the house, but as the
head of the house in a wider sense was also lord
and monarch of his people, the word came to mean
a lord or monarch ; and it is perhaps not altogether
an accident that we have no evidence of this in
Gaulish nomenclature, while it is well attested in
Kymric and Goidelic proper names : take the fol-
lowing : — Welsh — Catteyrn (Oattegirn, catoti-
GiRNi), Cyndeyrn (Kentegerni), Dutigirn, Euti-
girn, Gwrtheyrn (Vortigern), lUteyrn, Myllteyrn,
Rhydeyrn (Rutegyrn), Teyrn (Tegyrn), TeyruUuc,
Teyrnog (Ir. Tighearnach), Teyrnon ; from Corn-
wall we have tegeknomali, and in manuscript



(delwedd B60)


Tigerinomalum ; Breton — Maeltiern, Tiernmael ;
Irish — Eachtighearn, Tighearnan, Tighearnmas.

To the same Goidelo-Kymric period I would refer
the adoption by the Insular Celts of Druid ism,
which is probably to be traced to the race or races
who preceded the Celts in their possession of the
British Isles. Ctesar's words as to Druidism are
so well known that they need not be here quoted
at length. On the other hand, the Irish word for
druid, the equivalent of our derwydd is draoi,
genitive druadh, which in Irish literature mostly
means a magician or soothsayer, and is usually
rendered by magus in the lives of Irish saints
written in Latin. It has not been proved, as
pointed out by M. d'Arbois de Jubainville {Les
Celtes — Extrait de la Revue Archeologique : Paris,
1875), that Druidism found its way into Gaul
before 200 b.c. When it did get there, it was,
undoubtedly, through the Belgae, who had settled
in Britain: Caesar's words are significant (vi. 13)
— " Disciplina in Britannia reperta atque inde in
Galliam translata esse existimatur, et nunc, qui
diligentius earn rem cognoscere volunt, plerumque
illo discendi causa proficiscuntur."

As already pointed out, such items as these do
not amount to a demonstration of the correctness
of the classification here advocated ; but neither is
a demonstration necessary in order to give it a



(delwedd B60)


superiority over the one now in fashion. The
latter being shown to be founded on a misconcep-
tion, the former cannot but in the main be
admitted; and in any case it has the iraportant
consideration to recommend it, that it makes no
unnecessary postulates. A Celtic people speaking
one and the same language came from the Con-
tinent and settled in this island; sooner or later
some of them crossed over to Ireland and made
themselves a home there. The latter opinion is
countenanced, as far as they go, by some of the
names on both sides of the Irish Sea as given
in Ptolemy's Geography. Thus the Brigantes
occupied not only the North of England, but also
a part of Ireland: we have a Fayyavaiv uKpov in
the third of Carnarvonshire called Lleyn, and
Tafyyavoi located, as it is supposed, in what is now
called Clare. Possibly also Ptolemy's OvevUave!;,
in Forfarshire, belonged to the same tribe as his
Irish OvevvUvioi, or at least to a nearly related
tribe. Dr. Reeves in his edition of Adamnan's
Vita Sancti Columbus (Dublin, 1857) mentions,
p. 31, Inbher Uomnonn (in the map prefixed to
the work it is Inbher Domhnann), the old name of
the Malahide river, near Dublin ; also the Eirros
Domno of Adamnan's text, in Irish lorrus Domk-
nann, the barony of Erris in the county of Mayo,
which the Irish, according to his account, refer to




(delwedd B60)


the '' Fir Domhnann, Viri Damnonii, a section of
the Firbolgs." The Irish names here alluded to
suggest a connection with the Dumnonii of Devon,
Ptolemy's Aovuvovtoi, rather than with his Aafivovioi
of the North, or his Aafivoviov to kul Oxpivov aicpov,
supposed to be the Lizard, in Cornwall. Owing
to their being separated by an intervening sea,
there grew up between the Celts of Ireland and
their kindred in this country differences of dialect,
to which the probable adoption of their language
by races, whom they may have found in possession
of both islands, more or less materially contributed.
In the course of many centuries these differences
had become so many and such that they could no
longer be said to speak one language, but two
nearly related languages, Goidelic in Ireland, and
Kymric here. This is not altogether mere theory,
for all the most tangible differences between Welsh
and Irish can be assigned to various periods of
time posterior to the separation : this has already
been indicated in the case of a few of them, and
others will be dealt with as we proceed. Where
then is the necessity for supposing that the Celts
who took possession of the British Isles were even
then of two distinct nationalities, speaking two
distinct languages, and what was it that originally
determined that duality ?

Those who profess to be unable to believe that



(delwedd B60)


the Welsh and the Irish are nearly related, because
they find them unlike inr their national character
and habits, choose to forget how different the
circumstances were under which tl\ey have lived
from the days of Julius Csesar to our own. But
even so late as the reign of Elizabeth, their differ-
ence of history had hardly produced so marked a
difference of character as one might have expected.
Since then, however, the gulf has been consider-
ably widened. The Irish have had their '98, and
the bulk of them remain true to the Church of
Eome, while the Welsh have become Protestants,
and most of them have adopted the theological
views of Calvin, the force of whose influence, if we
look at it merely as a means of profoundly modify-
ing a people's character, and without regard to its
characteristics in other respects, cannot easily be

( 36 )



(delwedd B60)


"The initial changei^ are commonly the most perplexing feature of
the Welsh language to those who know it only imperfectly ; and
those who observe the rules by ear are seldom acquainted with the
rationale of their own faultless speech."— Chaeles Williams.

In spite of what was said in the former lecture,
you will perhaps think that, although the chief
differences between Welsh and Irish can be shown >
to have sprung up since the separation, the fact
of their springing up at all points to radical dif-
ference in the constitution of the vocal organs of
the Welsh and the Irish. It may, however, be
premised that this does not follow, as it is to be
borne in mind that the normal state of lan-
guage is that of change, and that the same end
may frequently be attained by different means.
The end here alluded to is not the ultimate end
of language, the expression of thought, but the
economy of labour in the articulation of words,
the exponents of thought. This, in default of a
better name, one may call its economical end.
This will appear plainer from a discussion of the



(delwedd B60)


so-called system of mutation of initial consonants
in "Welsh, and its counterpart in Irish, a subject
which, even apart from its relevancy to the ques-
tion how nearly Welsh, and Irish are related, has
strong claims on our consideration, though we run
the risk of only adding another chapter to the
mass of nonsense already written on it. The
fact is, our native grammarians, both Welsh
and Irish, look at it as at once the peculiarity
and the pride of Celtic phonology, and regard
it with the same air of mystery and wonder-
ment to which English and German gram-
marians occasionally give expression d propos
of the Teutonic ablautreihe or sing -sang -sung
system of vowel mutation obtaining in lan-
guages of that stock. In reality there is
nothing peculiar about either excepting the per-
sistency with which they have been carried out ;
and as to the amount of credit they respec-
tively reflect on the races which in the course
of ages unconsciously and cleverly pieced them
together, that is a matter on which opinion
seems to vary according to the writer's nation-

The following summary of the more common
mutations in Welsh and Irish will be found con-
venient as we go on : —



(delwedd B60)

ń38      ńLECTUEES ON WELSH PHILOLOGY.       Welsh.    Ikish.      Welsh.    IBISH.    c    g    ch     ń1    , lit  ^ rdd    \^'    t    d    th    r    p    b   - (gh)    ph   gli     ń1  r  1  r        b    dd = «   f = V    dh  bh    m    f = V    mh    CO    ch-    CO, c          t,t,    th    tt, t     ń70  nt     ń7c, 77h  nt, nnh      t    PP    ph or S    pp. p    mp    mp, xuiiiii    P      e            r.t    i Uh    cht     ń7g     ń77     ń77          nd    un    nu      y      mb    mm    mm    gg    cc, g'    c,^     ń1       Ih     ńLb    dd   bb    a, d'  pp,b'    t, d'  p, V      t     ńLit    i*     ń11     ń11     ń11    r    rth    Ir    Urh    Ir         ńLU^     ń1    rr    rrh, rh    IT      P    r P    rl    rU    rl      g    I [gtlijj    i^'    nl  nr    nil  nrh    nl  nr     Irish mutation, such as that of c into ch, or h  into hh (pronounced »), is commonly called aspira-  tion, and that whereby nt becomes t, or 7id nn, has  been more happily called eclipsis, while our own  grammarians have managed to include the Welsh  changes corresponding to both sets and others not  usual in Irish in the following triad : —     Radical.    Middle.    Nasal.    Aspirate.    c    g    ngh    ch    t    d    nh    th    P    b    mh    ph    g      Kg      d    dd    n      b    f    m       ń11;    • 1        m    f        rh    r          ńLECTUKE n.     



(delwedd B60)


This neat little scheme is fairly accurate in an
etymological sense, but it has not unfrequently
been assumed to have a phonological value, which
leads to mistakes, such as, for instance, the sup-
position that II is related to I in the same way as
t to d, and not as th to dd or nearly so. For our
present purpose the Welsh consonants may be
classified as follows : * —

Oral Coksonahts.

Nasal Consonants.


Spirants. ' '












ph or Łf









Here there are two things which require to be
clearly realised : the first is the difference between
a mute (otherwise called a stopped or explodent
consonant) and a spirant (otherwise called a pro-
duced or fricative consonant). Compare, for in-
stance, p and b with ph and v: in the former
two the breath is suddenly checked and stopped
by the lips being brought into contact with one
another, while in the latter two there is no com-

* Do rh, ngh, nh, mh consist of single consonants, or are they made
up of surd r, ng, n, m plus h, is a question I leave undecided ; the
latter view seems to suit Welsh phonology somewhat better than
the other.



(delwedd B60)


plete stoppage of it, since it is, so to say, allowed
to squeeze through without interruption. The
next is the distinction between surds (otherwise
called voiceless or pneumatic consonants) and
sonants (otherwise called voiced or phonetic con-
sonants), as, for instance, between p and 5, or
between ph and v : thus p and ph in the Celtic
languages imply simple breath, while h and v in-
volve not mere breath, but voice, which the former
produces by setting the vocal chords in vibration
during its passage through the larynx. It is
hardly necessary to state, in so many words, that
the vowels are both sonant and spirant, as they
are in fact almost pure voice more or less modified
in its passage through the mouth or nose.

Now one of the causes which bring about
changes in language is the tendency, ever quietly
asserting itself, to economise the labour of pronun-
ciation, and it is heterogeneous sounds brought
into immediate contact with one another, mutes
with spirants, or surds with sonants, that
form the hollows to be filled and the hills and
mountains to be lowered by the unreasoning lazi-
ness of speech : this levelling process is com-
monly called assimilation.

Let us now see how it will enable us to under-
stand the mutations of consonants in Welsh and
Irish : — Old Welsh ahal, ' an apple,' and aper^ ' a



(delwedd B60)


confluence, a stream,' became in later "Welsh afal
and aher respectively ; and why ? In ahal the h
was flanked by vowels, that is, a sonant mute by
sonant spirants ; and here both Welsh and Irish
took the same path, and reduced the mute into a
spirant, making aha into ava, written in Welsh
afa : in the latter we have a surd mute between
sonant spirants ; and as language proceeds by
degrees, and not by leaps or strides, it had the
choice of two courses, and only two : — it might
either reduce the surd mute into a sonant mute,
thus making aper into aher, or reduce it into a
surd spirant, which would give us apher. The
former has become the rule in Welsh and the
latter in Irish. But Irish in its later stages in-
dulges also in the Welsh mutation : thus such Old
Irish words as cet, ' hundred,' and cdic, ' five,'
are now c^ad and cuiff ; and so in other in-
stances where Old Irish c, t, (j>?) stood for no,
nt, (mp?).

Here you may ask how these changes, which
seem to have nothing to do with initial conso-
nants, have got to be known in Welsh grammar as
the mutations of initial consonants, or simply initial
mutations. The answer is not far to seek. The
action of assimilation in modern Celtic languages
is not confined to single words, but in certain cases,
which you learned when you were children, and



(delwedd B60)


which you will find enumerated in elementary
books on Welsh grammar, two words are taken
together so closely in speaking that, for the pur-
poses of phonology, they form as it were one, and
not two : thus the initial consonant of the second,
assuming it to have one, becomes a quasi-medial,
liable to the same changes as an ordinary medial.
For instance, though pen (Irish ceann) is head, we
say dy ben (Ir. do cheann), ' thy head,' and ei ben
(Ir. a cheann), ' his head.' Now these mutations
and the like are constantly recurring phenomena
in Welsh (and Irish) as now spoken and written,
and no writer on our grammar could overlook
them ; while to contrast aber with its older form
aper seldom occurred to them, and when it did,
they only found in the latter an orthographical
freak of the ancients ; and their ideas of the com-
parative immutability of their mother-tongue led
them tacitly to assume that aper was always pro-
nounced aber. Thus it was natural that they
should have called the changes in question initial
mutations, to which they ascribed a syntactical
rather than a purely phonetic origin.

That our grammarians, however, are not the
only class of writers who have failed to acquire
a correct idea of this kind of mutation, is
proved by the fact that it is the custom of
philologists to speak of it as though it were a



(delwedd B60)


property only of consonants flanked by vowels, or,
as they briefly term them, vowel-flanked conso-
nants — a description which would lead one to ex-
pect that the change could not go on when the
consonants are final, or come in contact with the
liquids I and r. Now it is remarkable that these
last are present in all the earliest attested cases
of this mutation, namely, in the following words
from the Oxford and Cambridge Glosses, together
with the Luxembourg Folio : — Dadl (for datl),
" concio," cedlinau (for cetlinau), ' to pursue,'
scribl (for scrip I), " scripulus," maurdluithruim
(for Tnaurtluitkruim), " multo vecte," ardren
(for artren), " pr^pugnis," riglion (for riclion),
"garrulis," cedlestneuiom (for cetlestneuiom),
" tabe." Thus the mutation in contact with one
of the liquids is the only kind known in the
earliest specimens of Old "Welsh : between vowels
it only began towards the close of that period in
the history of the language. The import of this
fact, translated into phonology, seems to be that
the liquids I and r have a greater power of assimi-
lation in Welsh than the vowels have. Suppose I
to stand for I or r, and jo for any mute consonant,
also a; for any quantity much greater than 1, then
you might roughly say that the tendency of the
language to reduce —



(delwedd B60)


][/>]! into l[i5]l = 2x,
a[p']\ into a[3]I = 2x — 1, and
a[/?]aintoa[5]a= 2x — 2.
These equations suggest another, namely, that
of a[j»] into a[6] =^ = x - \. Translate
this into a chronological form, and it means
that final mutes remained proof against muta-
tion after medial ones had been subjected to
it ; but does this agree with facts ? If you turn
to any tolerably well-written specimens of Med.
Welsh prose, such as most of the Mabinogion are,
you will find that it holds true in the case of c, t,
p : in fact, such forms as redec, goruc, dyfot, oet,
paraut, continually recur, but final p appears much
less frequently in them. Nay, it would seem that
traces of this had come down to William Sales-
bury's time ; for he says a propos of the letter c :
" Also other some there be that sound c as g, in
the last termination of a word : example, oc, coc,
Hoc : whych be most- commonly read og, cog, Hog "
(Ellis' Uarly English Pronunciation, p. 749).
This would bring us down into the middle of the
16th century. As to g, d, b, and m, they had
long before undergone the mutation in question,
whence it may be inferred that their power of
resistance was less than that of c, t, p. Thus
, it would seem that to achieve the nine mutations
forming the column headed ' Middle ' in the



(delwedd B60)


grammarians' table, it took the language at least
eight centuries. Strictly speaking, the process is
not yet complete ; for, in the Gwentian dialect,
Old Welsh t medial might be said to be still t,
as in oti {= ydyw), 'is,' ffetog (=: arphedog),
*an apron,' gatel (= gadael), 'to leave,' retws
(= rhedodd), ' ran,' and innumerable others, But
even here it cannot be said that no move has been
made towards the complete reduction of t into d ;
for the Grwentian t in the above words and the like
is not our ordinary t, but a t somewhat softened
towards d, a variety which I think I have also
heard from English peasants in Cheshire. So that,
after all, the Gwentian can only, be said to have
lagged behind the other dialects. This case, how-
ever is instructive as casting some light 'on the
question how t comes to be mutated, into d. Thus it
appears that Welsh t and d are only termini, between
which an indefinite number of stages have been gone
through, somewhat in the following order :— t, ti^

hj '3> • • • • 'n-lj ^ni ^ ^ni ®n-l> "3, <^j ^fj, a.

The varieties from t to t^-i inclusive would be
written if by a person writing from dictation, while
those from (4-i to c? would be written d: as to t„
and d^, he would hesitate between t and d; and
this no doubt is one reason why t and d were con-
founded in Med. Welsh, and even indifferently
written by the same persons in the same words.



(delwedd B60)


The same remarks, of course, apply to the other
surd mutes. It is needless to observe that this
kind of confusion could hardly have arisen had c,
t, p, been mutated into g, d, b, without any inter-
mediate steps. The view here advanced has, more-
over, the advantage of being in perfect keeping with
one of the most sacred dogmas of modern philology,
that all changes in language proceed by degrees.

By way of analogies in other languages, it will
be worth the while to mention just a sufficient
number of instances to show that mutation, in the
sense it has in Welsh grammar, is not peculiar to
our family of languages. In the first place, it may
be pointed out that in Sanskrit dsit + rdjd and
samyak + uktam become en phrase : dsid rdjd or
dsidrdjd, " erat rex," and samyaguktam, ' well said ; '
and so whenever a surd comes before an initial
sonant. In the interval between Latin and writ-
ten Spanish, mutation has regularly proceeded one
step, as in pueblo and trinidad from the Latin
populum and trinitatem : but, since the present
orthography, that is as far as concerns the conson-
ants, was established, it seems to have taken an-
other, as pueblo is pronounced with b like a labial v,
and trinidad with d as soft as our dd. Lastly,
Italian, according to Prince L.-L. Bonaparte,* dis-

* My attention was first called to this coincidence by a mention
in Ellia' Early English Pronunciation of Prince L.-L. Bonaparte's dis-
covery, which he has briefly given in his preface to II Vangdo di S.



(delwedd B60)


tinguislies a strong and a weak pronunciation of
the consonants, which are distributed in very much
the same way as the radical and reduced conson-
ants of Welsh, which we have been discussing.
So, in this respect, the pronunciation of Italian is
now in the same state as that of Welsh must have
been just before it had reduced c to ff, and so on.
Nor is this all : some of the Italian dialects have
gone as far as Welsh in this path of phonetic
decay, or even outstripped it. The most remark-
able is that of Sassari, in the island of Sardinia,
where, for instance, one says lu gori for Italian il
more — Welsh y galon, ' the heart ' (radical, cori,
calon) ; la derra for Ital. la terra — Welsh i dir, ' to
land ' (radical, terra, tir) ; and lu bobbulu for Ital.
il popolo — Welsh y bobl, ' the people ' (radical,
pobbulu, pobl) : a similar change takes place in the
case of radical g, d, b, s.

The second group of our mutations consists of
the reduction of yc, nt, mp into <yy^, nnh, mmk,
and of lyg, nd, mh into yy, nn, mm, respectively.
Let us begin with the latter three : in Mod.
Welsh they are written ng, nn (or n), mm (or m),

Matteo vdgarizzato in Dicdetto Sardo Sassarese dal Can. O. Spano
(London, 1866). The book is not easy to procure, and I am in-
debted to the Prince's kindness for a copy of it. Since then I have
incurred a similar obligation to Dr. Hugo Schuchardt of Halle, who
has written an elaborate article on the subject in the Romania.
There he discusses the consonants and their mutations much in the
same way as I have attempted in this lecture.



(delwedd B60)


and so- in Mod. Irish, excepting that, when a
quasi-medial is concerned, nn, mm, are represented
by n-d, m-b, in which the ofand b are not intended
to be heard. Thus it is hardly necessary to re-
mark that the assimilation is the same in both
languages ; however, it seems to have been neither
very common in 0. Irish, nor so inexorably carried
out in the subsequent stages of the language as in
Welsh, where we find it an all but accomplished
fact in our earliest manuscripts. One of the
latest Welsh instances of a medial complex ap-
parently free from its influence occurs in the
name Vendumagl-i on a stone inscribed in mixed
Eomano-British and Hiberno- Saxon characters of
the 6th, or more probably of the 7th, century:
later this name appears in the form Gwen-
fael. To this I will add two or three instances
more, which will suffice to convince you that what
we are discussing is more familiar to you than
you have, perhaps, anticipated : — annaearol, ' un-
earthly,' for an + daearol, ' earthly ; ' canm/ll
(pron. cannim/lT), 'a candle,' from Lat. candela;
am (pron. amm), ' about,' Ir. imm, im, represented
in 0. Gaulish by ambi, and in Greek by dfi(f)l ;
cam (pron. camm), 'crooked,' Ir. camm, cam, for
camb — as in the 0. Gaulish Cambodunum. The
same thing also happens when the mute is a
quasi-medial, as, for example, after the proclitic



(delwedd B60)


preposition yw, 'in,' as when we say yw Nimhych,
' in Denbigh,' yn ninas Dafydd, ' in the city of
David,' for yn + Dimbych and yn + dinas : so in
other cases too numerous to mention.

To return to the other three, they are, after
undergoing eclipsis, as Irish grammarians call
it, written in Mod. Welsh ngh, nnh, mmk, which
imply a process that requires some explanation.
The veteran phonologist, Mr. Alexander J. Ellis,
who has written extensively and elaborately on
Early English Pronunciation, considers that the
n in the English word tent is partially assimilated
to the following mute, and that it becomes a surd
which he would write nh: thus he would repre-
sent tent as pronounced tennht, and similarly
tempt, sink, as temmht, siqqhk — his q means the
sound of ny in sing, for which I have made use of
y. It is hardly probable, however, that any ordi-
nary orthography would take cognizance of the
difference between surd and sonant nasals in the
positions here indicated, and I am inclined to
think that the Welsh of old who wrote hanther,
now hanner, 'half,' and pimphet, now pummed,
' fifth,' meant something more than this. As the
spirants th, ph, are out of the question, it is not
improbable that nth, mph, were intended to be
pronounced nkt^h, mhp^h, that is, the complexes
nt, mp, were to be aspirated, which we may express




(delwedd B60)


by -writing them [nt'jk, [mp'Jk, respectively. So
far Irish may possibly have proceeded on the
same course as Welsh, but no further; for the
next step it takes is to allow the nasal, whether it
was surd or sonant, to disappear, leaving the pre-
ceding vowel in certain cases — possibly only when
it had the tone — lengthened to preserve the quan-
tity of the syllable. Thus it converts such a form
as dent into ddt, ' a tooth,' that is del, for the
Irish use the acute accent to indicate quantity.
Now det is in Welsh dant, which is free from the
eclipsis, but not so its derivative dannheddog,
' toothed.' Here not only has nt become \nt\h,
but the nasal which began to be assimilated by
the oral consonant eventually vanquished the
latter and completely assimilated it to itself in
its altered condition, so that for \nt\h we get
\nn\h^ that is, in our ordinary orthography, nnh.
Other instances, such as tymmhor, ' a season,'
plural tymmkorau, from the Latin tempus, tem-
por-is, and annheilwng, ' unworthy,' for an +
teilwng, ' worthy,' are so common that I need not
mention more of them ; nor is it requisite to dwell
on the similar eclipsis of quasi-medial c, t, p, as,
for instance, in yn nghwsg (pron. y'nghwsg),
' asleep,' yn Nhynyn, ' at Towyn,' for yn + cwsg
and yn + Tywyn. But why, to revert to one of
the instances just mentioned, should tymmhor,
which seems to have been preceded by tylmplhor



(delwedd B60)


or tymkp'hor, for Latin tempor-, have taken the
place of that form ? that is, why should p'h have
yielded its place to mh ? Here, as before, the
answer must be sought in the tendency of lan-
guage to lessen by assimilation the labour of
utterance. Thus, in the case before ns, the jo'A
(oral, mute, surd) standfe between mh (nasal,
spirant, surd) and the vowel o (oral, spirant,
sonant) : so it seems perfectly intelligible that
the language, proceeding by degrees, should
replace ph by a surd spirant ; but that would
leave us in the dilemma of having to decide be-
tween the nasal spirant, mh and the oral spirant
ph (== ff), that is, between tymmhor and tymphor.
This, however, an unerring instinct does for us in
favour oi tymmhor * the reason probably being, that,
as we have already seen in another case, the assimi-
lative power of a consonant is greater than that of
a vowel, that is, in this instance, of m than of o.

Thus far we have traced <yc, nt, mp, through two
stages of modification : sometimes, however, the
language goes a step or two further, and in cer-

* Substitute for the vowel I or r, and the reTerse takes place,
the oral consonants having, it would seem, more assimilative force
than the nasal. The instances are not very numerous — I may
mention cetkr, ' a nail, a spike,' for centhr, Breton Jeentr, borrowed
from the Latin centrum, and cathl, ' a song,' for canthl, whence the
0. Welsh cenihliat, " canorus ; " the gloss occurs in the Juvencus
Codex on the words Dauida canorum, and would now be cethliad:
the disappearance of the nasal is a later step, which has nothing to
do with the assimilation in question.



(delwedd B60)


tain cases even confounds the representatives of
these complexes with those of 7^, nd, mb : — (1.)
In Mod. Welsh we find it difficult to double a
consonant not immediately following the tone-
vowel, consequently such words as dannheddog and
annheilwng become danheddog and anheilwng in
pronunciation. Similarly we neither speak nor
write fyn nhad, fym mhen, but always fy nhad,
'my father,' fy mhen, 'my head,' for fyn + tad
and fyn + pen, the possessive pronoun being a
proclitic, which never has the tone. It must, by
the way, be explained, that although in book
Welsh the word is written fy, even before vowels,
as in_^ enw, ' my name,' liable to become in North
Wales y^MW, and so in other cases, the old form
of it was min, which is still duly represented in
South Wales hj fyn — in North Cardiganshire it
sometimes becomes fyng, like pring for prin,
' scarce ' — as va.fyn enw, ' my name,'_/y?j oen, ' my
\&,mb,^ fyn arian, ' my money,' and the like : it is
this full iorvafyn that must be considered in the
eclipsis. Add. to the foregoing the case of r^g, nd,
mb, which is similar. Thus we say fy nydd, ' my
da,y,,^fy mramd, 'my brother,' ^at fyn nydd, fym,
mramd, ioxfyn + dydd&ndifym + brawd: similarly,
we say saith mlynedd, ' seven years,' for saitKn
+ blynedd, Irish seacht m-bliadhna (pron. seacht
mliadhnd) ; for saith is one of our numerals which
originally ended in n, matched in Latin by the m



(delwedd B60)


of septem, novem, decern ; and such a phrase as saith
rdynedd is an interesting instance of a fact re-
maining long after one of its factors is clean gone.
Occasionally the nasal is also simplified when it
happens to be medial, as in ymenyn, 'butter,' for
which one might have expected ymmenyn for ymb-
en-yn : the Breton forms are amanenn, amann,
and the Irish imb, imm, im, all from the root
AKGV, whence also Lat. unguo, ' I smear or be-
smear,' AUemanic anko, ancho, ' butter.'

(2.) The surd is liable to become a sonant unless
it comes immediately before the tone- vowel : thus
such words as dnghlod, ' disrepute,' dmmkeu, ' to
doubt,' tymmkor, ' a season,' are sometimes pro-
nounced dnylod, dmmeu, tymmor ; that is, a second
process of assimilation has taken place in them ;
but it is prevented by the position of the tone in
ammMuaeth, ' doubt,' and tymmhorau, ' seasons.'
.In words such as the following no trace of the
surd is to be found : — cdnnoedd, ' hundreds,' ddn-
nedd, 'teeth,' which is followed by the North-
walian pronunciation of dannMddog as daniddog,
tdnnau, ' chords,' trSngu, ' to expire,' from trangc,
' death,' and many others. As to such words as
ugain, ' twenty,' and drjan, ' money,' for ugaint
and arjant, they seem to be instances of the retreat
of the accent from the ultima to the penultima,
accompanied by the reduction and the simplifica-
tion of the nasal : a similar remark would seem to



(delwedd B60)


apply to the colloquial form of the third person
plural of verbs, as when ydynt, ' sunt,' clywsant,
' audiverunt,' rhedent, ' currehant,' are made into
ydyn, clymson, rheden, a pronunciation which no
one would, however, use when reading in public.
The case of the woTdyminnydd, ' brain,' is different
and somewhat exceptional : as the Breton is em-
penn, and the Irish inchinn, genit. inchinne (com-
pare the Greek iyKecpaXo^) , we might expect it to
be in Welsh ymmhennydd or ymhennydd for yn~
penn-ydd. The explanation would seem to be that
the word was formerly accented ymennydd.

It has already been hinted that y, rf, h, have less
power of resistance than <?, #, p : this is confirmed
by the history of the modifications we are now
discussing. Thus, while the eclipsis has in the
case of the former three been approximately as-
cribed to the 7th century, few instances of its
affecting the latter are to be found in the range of
0. Welsh, but as we pass on to Mod. Welsh we
find it far from unusual in a manuscript which
Aneuriu Owen supposed to be of the 12th cen-
tury. I allude to the Venedotian version of the
Laws of Wales. Later, in the Mabinogion, we
have such forms as cyghor (pron. cy^yhor), 'counsel,
council,' amherawdyr, ' emperor,' from Lat. im-
perator, side by side with ympen, 'in the head,'
ygkairllion, ' at Caerleon,' which are now pronounced
ymhenn and y^haerlUon ; and so in other cases. In



(delwedd B60)


instances of this kind a disinclination to obscure
what may be called the dictionary form of words
must be regarded as having for a time stemmed
the current of phonetic decay. Still later Sales-
bury is found indulging in such combinations as
yn-pell, ' far,' and yn-carchar, ' in prison ; ' but
according to his own account the mutes following
n were dead letters, which he only meant to appeal
to the eye : it is easier to forgive him this than
such freaks of fancy as vy-tat, vy-bot^ for vy nhat,
' my father,' and vy mot, ' my being,' which do
much to detract from the phonological value of his
writings. Perhaps one of the last conquests which
eclipsis has made in Welsh occurs in our colloquial
ynkwy, ynkm, nkw, for the written A/vyntAwy, that
is, Awynt-Awy, ' they.' For I need hardly say that
one or more words have already been cited which
may have reminded you that those conquests have
hitherto not been complete ; — whether that would
continue to apply to them, supposing the language
to live long enough, is a question which it would
not be easy now to answer. In the first part of
this lecture it was noticed that the reduction of
a[j»]a into a[5]a took place earlier than that of
a[jo] into a[^] : the parallels to these in the case
of eclipsis are the reduction of m&lp7i]a into
mklmk^a and that of mAlp'A'] into mA['mA'], that is
in pronunciation, as this concerns a final conson-
ant, mm, now commonly written m. Now it is



(delwedd B60)


mainly words which come under this formula that
have successfully resisted eclipsis, such, for instance,
as the following : — dant, ' a tooth,' plural dannedd;
hynt, 'a journey,' 0. Ir. sit; 'pump, 'five,' 0. Ir.
coic ; tant, ' a chord,' plural tannau, 0. Ir. tdt, Mod.
Ir, teud; meddiant, ' possession,' plural meddian-
nau. To these may be added cant, ' a hundred,'
plural cannoedd, 0. Ir. cet, Mod. cdad, which forms
a sort of compromise between the rule and the
exception ; for we saj pedwar cant, ' four hundred,'
but can (pron. cann) erw, ' a hundred acres,' and
can.ych, 'a hundred oxen.'

Now that the ground which this part of our
inquiry should cover has been rapidly run over, it
may be added that there is nothing in eclipsis
which may be regarded as peculiar to the Celtic
languages ; but I will only cite from other languages
just a sufficient number of analogous instances to
indicate some of the quarters where more may be
found, (a.) You may have wondered how such
English words as the following, now pronounced
dumm, lamm, clime, came to be written dumb, land),
climb: the answer of course is that the b in them
was formerly pronounced, and that this is merely a
case of the spelling lagging behind the pronuncia-
tion — littera scripta manet. To this class of words
may be added the modern rcoodbine, which at an
earlier stage of the language was written wudubind;



(delwedd B60)


and, to come down to our own daj'^, all of you have
heard London called Lunnun. Beyond the Tweed
this and more of the kind may be considered
classic : witness the following stanza from Burns'
Five Carlins ! —

" Then neist came in a sodger youth,
And spak wi' modest grace,
An' he wad gae to Lon'on town
If sae their pleasure was. "

Here may also be mentioned, that there are
German dialects which habitually use kinner,
wunner, mermen, unner, branmvm, for the book-
forms kinder, wunder, wenden, unter, branntmein.
Similarly in 0. Norse bann and lann are found for
band and land, not to mention the common reduc-
tion of 7^ into nn as in finna, ' to find,' annar,
'other' (German under), munnr, 'mouth' (Ger.
mund) , and the like. (b. ) Diez in his grammar of the
Romance languages supplies a variety of instances
in point, such as the following : — Siciii&n, abbunnari,
' abbundare,' accenniri, ' accendere ; ' Neapolitan,
chiommo, 'plumbum,' munno, 'mundus.' And it
is perhaps by assimilation that nd, nt final have
become n in Provencal, as in gran, ' grandis,' joreow,
' profundus,' fron, ' frons, frontis,' den, ' dens,
dentis,' joven, 'juventus.' (c.) So far I have
failed to discover an exact parallel to the Welsh
eclipsis of c, t,p, leaving the nasals in a su^d state
as in our stock instance tymmhor from tempus.



(delwedd B60)


temporis ; but this is probably to be attributed to
my very limited acquaintance with the exact
pronunciation of other languages. It would not,
however, be altogether irrelevant here to mention
Mr. Ellis' account of the sound of n, for instance,
in the word tent, which he regards as pronounced
tenht or tennht, and to add that he further finds
that in Icelandic n coming after ^ or ^ is also made
into nh, as in vatn, ' water,' regin, ' rain,' pro-
nounced vatnA and regknh respectively. Now there
can be no doubt that at one time English kn also
was, provincially or generally, pronounced knh; for
when the k ceased to be heard in such words as
knave, knee, know, the nh still remained, a point
amply proved by Cooper, who published, in 1685,
a work entitled Grammatica Linguce Anglicance,
from which Mr. Ellis cites no fewer than five
passages giving the then English pronunciation of
kn as hn. This kn, which we are wont to write
nh, and Cooper mentions in company with zh, wh,
sh, th, as having no place in the alphabet, found its
way into Wales, nor has it to this day quite dis-
appeared from our pronunciation of English. When
I was a boy, our schoolmasters in Cardiganshire
prided themselves on the many things they nhew,
and favoured the boys who strove to benefit by
their superior nhonledge, but as to the young nhaxes
who preferred idling, they had their laziness liter-



(delwedd B60)


ally nhocked out of them in no pleasant manner :
in fact, there are Welshmen not a few still living
who have never lost the nhowledge thus nhocked into
them when they were boys.

The next mutations to be noticed, in the order
given in the table we have been following, that is,
if we reserve It for a special mention, and omit Ip
for want of sure instances, are those of Ic, re, rt,
rp, into Ich, rch, rth, rph (or rff), as in the follow-
ing words : — golehi, ' to wash,' 0. Irish /olcaim,
" humecto,lavo; " march, 'a steed,' whence mwrehog,
' a knight,' Ir. mareaeh ; nerth, ' strength,' whence
nerth/hwr, ' powerful,' 0. Ir. nertmar, Gaul. JS'er-
tomarus ; corff, ' a body,' plural cyrff and corffo-
roedd, Ir. corp (Lat. corpus, corporis) ; gorpken,
' to finish,' from pen, ' head, end,' with the prefix
gor. The formula of the reduction in these words
and the like is not that of r\_p']a but of rh\_p'\a
into rh\_ph']a, that is, for instance, the Latin corpus
was, in Welsh mouths, corhpus, with p (surd,
mute) between rh (surd, spirant) and u (sonant,
spirant), so that under the combined influence of
its two neighbour-sounds it had to be changed into
ph (/), which gives us corff and not corb, as might
be expected were corpus to be treated as such and
not as corhpus. Even now, if I am not mistaken,
the liquids in corff and golehi are not quite the
sonants r and I, but rather rh and Ih ; or, perhaps,



(delwedd B60)


it would be more correct to say that they begin as
sonants and end as surds, to be timed ^-^^ and
ijwih respectively. Thus following Mr, Ellis' palaao-
type representation of tent as tennAt, we might
say that these words are pronounced corrhff and
gol-lhcki. When the spirants ch, th, ph, began to
take the places of the corresponding mutes in the
positions here indicated, it would now be hard to
say ; however, our earliest specimens, scanty as
they are, of 0. Welsh of the 9th century exhibit
them on much the same footing in the language
then as now. It is true that occasionally c, t, p,
are to be met with for ch, th, ph, but that is pro-
bably rather the result of carelessness in writing
than of any uncertainty in the pronunciation.
This phonetic change is not a very common one
in European languages ; but we seem to have an
implied instance of it in the Sassarese dialect in
such words as baj,ca (Ital. barca) and alchi (Ital.
archi, with the ch, as usual in Italian, standing for
the sound of k) : the present pronunciation is given
by Prince L.-L. Bonaparte as hw^Qfo, and a-^, so
that the intermediate stage can hardly but have
been hab^ and alyi or bar-^a and aryi.

The next mutations in the table are those of rd,
lb, rh, into rdd, If, rf, to which may be added those
of Im, rm, into If, rf. They need not be here dwelt
upon, as the same explanation applies to them as



(delwedd B60)


to vowel-flanked consonants and others mentioned
at the outset. But as to Ig, rg, it is to be noticed
that even in 0. Welsh they had the sound of Igh,
rgh, with gh sounded as the sonant spirant which
may sometimes be heard in such German words as
liegen, ' to lie,'i and regen, ' rain.' In the Oxford
Glosses on Ovid's Art of Love we have this once
written gh, namely in helghati, " venare," that is,
helgha ti, ' do thou , hunt ; ' but in the Cambridge
Glosses on Martianus Capella we have it written ch
in the verbal noun in the phrase in helcha, ' in
venando/ now.yw hela, ' in hunting :' compare the
Irish seilg, ' a hunting, venison.' Probably the
sound was the same even where g continued to be
written, as in 0. Welsh colginn, ' aristum,' now
coli/n, ' a sting.' The next step was to omit the
consonant altogether, as in the last-mentioned in-
stance, or else to change it into_/ as in kelghati,
now helja di ; and in such words as arjan, 'silver,'
0. Welsh argant, Breton arc'hant, 0. Ir. argat,
now airgead, from Latin argentum, tarjan, ' a shield,'
from 0. English targe, genitive targan ; to which
may be added proper names in gen, such as 0.
Welsh Morgen, TJrbgen, later Morjen, Urjen.

Next in order come ch, th, ph, for cc, tt, pp re-
spectively, as in sack ' a sack, from Latin saccus,
saeth ' an arrow,' from sagitta, and cyff ' a stump,
trunk, stem,' from cippus. The same thing hap-



(delwedd B60)


pens iu the case of quasi-medials, as, for instance,
when we use ac ' and,' tri ' three,' which stands
for an earlier tris, or ei 'her,' which originally
also ended in s, as in the following examples : —
ci a chath ' a dog and a cat,' for ci ac + cath; ty a
than ' house and fire,' for ty at + tan = ti/ ac + tan ;
tri phen ' three heads,' for trip + pen = tris +
pen ; ei Must ' her ear,' for eic + dust = eis +
dust ; whereas ' his ear ' would be ei glust, because
ei masculine originally ended in a vowel — the
Sanskrit for ejus is asya ' his,' asyds ' her.' This
mutation, so common in Welsh, to which I have
hitherto failed to find a parallel elsewhere, is pro-
bably to be explained as follows : — Take for instance
the Latin word cippus, which the Welsh borrowed
into their own language. Here the vowels i and u
are separated by two /»'s, whereof the one is implo-
sive, or formed when the lips are brought together,
and the other explosive, or formed when the con-
tact ceases. Now the assimilative force of the
vowels would tend to reduce the word to cibbus or
cipkphUs. But the double consonant generally also
implies a more violent ejection of air from the
lungs than is usual in the case of a single one, a
circumstance which is directly antagonistic to any
reduction in the direction of jo to 6 ; so dbbus is
ruled out of the field. Of course, in the case of
our supposed form dphphus, the two ph's being



(delwedd B60)


continuous sounds, could not fail to merge them-
selves into one, that is to say, if they were not to
be so regarded from the first. In either case the
result would be cipkus or ciffus, whence our cyff.
Then as to the time when this mutation became
the rule, that may be determined between certain
wide limits. It is an accomplished fact in the
9th century, whereas about the middle of the
6th century a Continental writer speaks of our
crwth as " chrotta Britanna." So it may be
ascribed to the 7th or the 8th century, proba-
bly the former, for which our inscriptional evi-
dence seems to make: an Anglesey tombstone bears
the name Decceti, while another, in Devonshire, in
letters tending to the Hiberno- Saxon style, gives
it the form Decketi. Still more instructive is an
inscription from Carmarthenshire which mentions
a man called Lunarlc^hi Cocci, in letters which
can hardly be earlier than the middle of the 6th
century. This last clearly shows that re had be-
come rch before cc had yielded ch as in coch ' red,'
the modern representative of cocc-i ; a fact which
is quite in harmony with what has already been
said as to the relative force of vowels and con-
sonants for assimilation.

The transition of such a word as cippus into
aphpkus or ciphph would lead one to expect fruc-
tus to have become in Welsh, in the first instance,



(delwedd B60)


fruchthus or fruchth, but we have no evidence
whatever for such a form. In the earliest 0.
Welsh we have ith in the place of chth, and, ac-
cordingly, /rz^zi^A, now ffrwyth, ' fruit,' for fructus;
and so in native words, such as wytk, ' eight,' for
oct, Ir. oeht, Lat. octo ; rhaith, ' law,' for reet-, Ir.
reckt, Lat. rectum, Eng. riffkt. Did cktk become
ith without any intermediate stage of pronuncia-
tion ? That is hardly probable, and the next
thing is to suppose the steps to have been ct, ckth,
ghth, ith, or rather ct, cht, chth, ghth, ith, as the
Irish equivalent is still written cM, though the
pronunciation, it is true, approaches chth or chtth.
And it is not improbable that cht dates from the
period of Goidelo-Kymric unity, if not earlier; and
it is to be noticed that, as ht, cht, or ght, it is com-
mon to the Teutonic languages, where it would
accordingly seem to date before their separation
from one another : take for instance the English
word might, formerly written meaht, miht, G-er.
macht, Gothic mahts. Then, in the next place,
as to the transition of cht into chth, it is just what
the analogy oi rth, rch, for rt, re, would lead one
to expect in Welsh ; but a more questionable step
is the softening, here supposed, of chth into ghth.
However, the pronunciation offers no difficulty, as
it is easy to begin the gh as a sonant spirant and
to finish the th as a surd one ; in point of assimi-



(delwedd B60)

ńLECTUEE 11. 65

lation, such a syllable as acht offers in its ch a
compromise between the a and the t. Moreover,
English orthography seems to have registered an
analogous process in such words as night, which
was formerly written neaht, naht, niht, then nigt
and night. The gh was sounded in English in
William Salesbury's time, who describes it as
softer than Welsh ch, but otherwise of the same
character. The change of spelling from h to gh
was preparatory in some of the instances to its
ceasing altogether to have the power of a conso-
nant, which happened with the same result as in
Welsh. Take again the word night with its short
i (as in pin) lengthened eventually at the expense
of the gh into I (as in Welsh, or ee in the English
beech) — the subsequent diphthongisation of that I
into the ei of our own day, which permits our
writing night in Welsh spelling as Tieit, does not
concern us here — and compare the Welsh word
hrlth, feminine braith, ' spotted, party-coloured ; '
brith stands for a much earlier brlct, which may
be supposed to have successively become bricht,
brichth, brighth, brith; while the feminine stands
for brictd, which would have to pass through the
stages brichta, brechtha, breghtha, breith, on its
way to our present Welsh braith. The presence
of an i for the first consonant in the combination
in question is common to Welsh with French, as



(delwedd B60)


in the Old Welsh fruith, now ffrnyth, French,
fruit, and so in other instances, a coincidence
which the advocates of the Gallo-British theory
should make the best of ; but as words borrowed
into Welsh from Latin follow the same rule in
this as native ones, it is probable that chth, ghth,
date after the Eomans came to our shores, and the
only inscription bearing on this point seems to
favour that supposition, as far at least as con-
cerns ghth. It comes from Pembrokeshire, and is
in letters which may, perhaps, be assigned to the
latter part of the 6th century: they can hardly
be much earlier. The reading seems to be Nog-
tivis Fill Demeti ; the Ogam differs, but it cer-
tainly begins with nogt, which I take to mean
noghth rather than nogkt, as I fail to see how the
latter form could have arisen : noghth would be the
prototype of one of the words which have the.
form noeth in Mod. Welsh ; that is to say, noeth,
' naked,' Ir. nocht, and noeth, ' night,' as in he-
noeth, ' to-night,' Mod. Ir. anocht. By gk is here
meant the same sound which yielded 7 in helja and
Morjen already mentioned, and which, as the con-
tinuator of g followed by I, r, or n, is replaced in
Mod. Welsh by the vowel e in such words as Mael
for 0. Welsh mail = magi, as in Grildas' Maglo-
cuni, aer, ' a battle,' 0. Welsh, air = agr-, of the
same origin as the Greek aypa, 'a catching, hunt-



(delwedd B60)


ing, the chase,' and oen, ' a lamb,' 0. Welsh, oin =
ogn-, of the same origin as Latin agnus. Irish is
satisfied with merely lengthening the vowels by-
way of compensation, so that the foregoing words
assume in that language the form Mdl, dr^ and
uan = on. These guesses, which cannot seem less
satisfactory to you than they do to me, would look
incomplete without a mention of pi; but as pt is
supposed to have been changed at a very early date
into ct, it has no history of its own. Thus our
saitk, ' seven,' formerly seitk, is regarded as the
direct representative of a Goidelo-Kymric seckt or
sect for an Aryan saptan, which is rendered pro-
bable by the Irish form, which is now .seacht,
formerly secM. And it is worthy of notice that
the only Latin loan-word with pt has been treated
in Welsh differently from those with ct : I allude
to pregetk, ' a sermon,' from. prceceptttm, 'a maxim,
rule, injunction, doctrine ' — compare also TrAipki,
' Egypt,' for 'H A'lyvTrTo<}.

We have not yet done with the table we set out
with : there still remain the items in Italics.
Instances have been noticed of the reduction of c,
t, p, into g, d, 6, but" now we have to deal with
changes which seem to take the other direction, as
when gg becomes cc and the like : this kind of
mutation may, in default of a more appropriate
term, be called provection. But when c, for



(delwedd B60)


instance, is reduced in. Welsh to g, we know that
to be an instance of assimilation tending to lighten
the labour of articulation; however, it is not to
be assumed that provection is a kind of dissimila-
tion to increase it. Let us begin with the more
palpable cases in point : what makes it so difficult
to teach a Welshman not to make the English
words hag, pod, tiib, into back, pot, tup, or to get an
Englishman to pronounce the word eisteddfod cor-
rectly as eistehvod, and not as eistethphod? It cannot
be that pod is made into pot because the o is fol-
lowed by a mute or a stopped consonant, for t and
d are the same in that respect ; and in the other
case th and dd are both spirants or continued con-
sonants. Thus it is clear that these changes do not
depend on any of the qualities serving as a basis
for the classification of consonants into mutes and
spirants into surds and sonants. Another glance
at the table will show that, when provection takes
place, more consonants than one are concerned.
Now it happens almost uniformly in Welsh, that
when an accented vowel is followed by a combina-
tion of consonants, it has a closed pronunciation,
which implies a hasty and forced ejection of air
from the lungs. This high pressure, so to say, is
not favourable to the pronunciation of such con-
sonants as g, d, b, dd,f, as they require the organs
of speech to be brought together much more gently



(delwedd B60)


and slowly than in the case of the corresponding
snrds. Hence it is clear that when a Welshman
makes hag into hack, or an Englishman eisteddfod
into eistethphod, these are cases of assimilation
based on a third principle, the force of the vowels,
and, in the instances before us, the assimilation
distinctly amounts to the substitution of an easier
for a harder pronunciation.

It is hardly necessary to state that the use made
of provection is only sporadic in Welsh as com-
pared with the other kinds of assimilation and
their far-reaching effects on the words of the
language. In Irish, however, it plays a con-
siderably more important part, whence another
divergence between the two languages, especially
in words which, in 0. Welsh and 0. Irish, con-
tained the combinations lb, rb, rd, seeing that in
later Welsh they are If, rf rdd, and in Irish lb, rb,
rd, or even Ip, rp, rt. Thus the 0. Welsh gilbin
becomes gylfin, ' a bird's bill or beak,' while the
Irish is gulha, which also occurs with a p instead
of h; and the Latin or do appears in Welsh as urdd,
' an order,' and in Irish as ordd or ort, genitive
uirdd or uirt. It may not be wholly devoid of
interest to you to find that there are cases of
provection in English in such forms, for instance,
as the perfects meant, 0. English mcende, mende ;
dreamt, 0. Eng. dremde ; dealt, 0. Eng. dmlde.



(delwedd B60)


delde; felt, 0. Eng. felde, felte, to which may be
added others such as built, 0. Eng. bulde, and
bent, 0. Eng. bende. The same thing happens
when the ending ard becomes art as in braggart,
sweetheart, a change which invariably takes place
in Mod. Welsh when words of this category are
borrowed, as, for instance, in godart, 'a kind of
cup,' sowgart, ' a riding habit,' llempart, Rhisiart,
from goddard, safeguard, leopard, Richard.

But, to proceed to instances of a more respect-
able antiquity, we come to gg, dd, bb, yielding
mutes : in order to avoid confusion they must be
treated as belonging to two strata of different
dates. The later of them belongs to Mediseval and
Modern "Welsh, and dates after most of the reduc-
tions already discussed had taken place, as, for
instance, in such words as these : cyttuno, ' to agree,
to bargain,' for cyd + duno, ' to unite, agree ; '
yspytty, ' a hospice,' for yspyd + dy (for ty, ' a
house ') ; Hetty, ' lodgings, an inn,' for lied + dy.
Here it is to be observed that when the tone falls
on the vowel immediately preceding the mutes in
question, the vowel is shortened and forced while
the mute is doubled ; but as soon as the tone
shifts, the vowel is slackened and the mute
simplified. However, it is usual to write lletty
' lodgings ' and llettya ' to lodge,' or else llety and
lletya; but neither orthography is accurate and



(delwedd B60)


consistent, for the •words being accented on the
penultima as usual, are pronounced Hetty and lletya.
This would perhaps be most readily indicated for
the benefit of strangers desirous of learning our
language by writing lUty and lletya. Similarly in
cases of assimilation we should have to write, for
instance, atebodd, ' respondit,' and dteb, ' respon-
dere,' for ad plus keb as in gohebu, ' to correspond
by letter ; ' in 0. "Welsh it is hep, ' quoth,' for a
European saqv-, whence the English say, German
• sagen, and the Lithuanian atsakyti, which is all but
bodily equivalent to our dteb.

The other stratum of instances alluded to be-
longs to 0. "Welsh, and they are, as might be ex-
pected, few in number. Apertk, now abertk, ' a
sacrifice, an offering,' would seem to be one, as it
admits of being analysed into (ap-pertk for) ab-
herth = ad-berth : the 0. Irish forms are edbart,
edpa^t, id-part (Zeuss^, p. 869), all from the root
ber, the Celtic equivalent oifer, in Lat.yer-o, Greek
<j)ep-a), ' I bear.' The analysis of the Old "Welsh
aper, now aber, ' the mouth of a river,' would dis-
close the same root, if one is right in understand-
ing the word originally to mean the volume of
water which a river bears or brings into the sea or
into another river. Compare Umbrian arfert-ur
(for ad-fert-ur), 'allator, oblator,' and arferia,
glossed by Festus "aqua qute inferis libabatur."



(delwedd B60)


To the working of the same principle in 0. Welsh
we are probably to trace apati, for abhati, in a
Latin inscription in Hiberno-Saxon characters on
a stone at Llantwit in South Wales. Similarly
Welsh cred-u, ' to believe/ for an earlier cret-u, 0.
Irish creitem, ' faith,' neither of which seems to be
derivable from Latin cred-o, is rather to be com-
pared with Sanskrit graddhd, ' trusting, faithful,'
^addhdna, ' faith,' graddMtaw/a= Welsh credadwy,
' to be believed.' We may probably assume that
aperth stands for an earlier apperth {=ahherth
^ adberth), and the conclusion seems natural,
that the simplification of the mute implies that
the accent was on the ultima: unfortunately we
cannot be said to know much about its position in
0. Welsh. However, the fact that aperth, for in-
stance, was pronounced aperth and not apperth in
the latter part of the 0. Welsh period is rendered
certain by its further reduction in later Welsh. into
aberth: so with the other instances.

Before leaving this point, you may wish to know
if anything corresponding happens in the case of
quasi-medials, that is, if we have parallels to the
phrases already mentioned, ci a chath, ty a than, and
ei chlust. There are such, and the following will
do as instances : tri gair, ' three words,' ceiniog a
dimai, ' a penny and a halfpenny,' ei bara, ' her
bread.' These might at first sight seem to be



(delwedd B60)


hardly in point, the forms to be expected being
tri cair, ceiniog a timai, ei para ; however, looking
at the actual ones, you will observe that the lan-
guage has not set out from tri gair, a dimai, ei
bara, for in that case we should now have by re-
duction tri air, a ddimai, ei far a — this last does
occur, but it means ' his bread,' and not ' her
bread.' The fact is, tri gair, for instance, with a g
that resists reduction, stands for trig + gair for an
earlier tris + gair. It is this kind of strengthened
g that has been entered in the table as g. A simi-
lar remark applies to d' and b'.

We now pass to the consideration of It and Id,
as to the former of which, it is possible that It, in
the first instance, became Iht by assimilation ; but
Ih, though a surd, is not the sound ,we write U,
which roughly speaking stands to I as_/ tot;, or th
in ' thin ' to th in ' this.' What is the exact rela-
tion in which our II stands to Ik ? would a change
from Ih into II be a case of provection, or is II due
altogether to the influence of the t following it ?
These are questions which I must leave in the
hands of those who make the physiology of speech
their special study. The combination Id also
yields lit, for the d in melldith, ' a curse,' and
melldigo, ' to curse,' from the Latin maledictio and
maledicere, is merely historical, the pronunciation
being melltith and melltigo ; nor does anybody, so



(delwedd B60)


far as I can remember, write swlld and cysylldu for
srclli, ' a shilling,' and cysylltu, ' to join, to con-
nect,' as it is not very generally known that these
two words are borrowed from the Latin soldus or
solidus and. consolidare. In the change of Id into
lit, language probably proceeded, as usual, by de-
grees : in the first instance Id became It by pro-
vectiou, which, by the way, is shared by Bretou,
for it is from It it must have arrived at the
vocalised ut, ot, which it opposes to our lit. The
next step was to make It, Iht, into lit ; bo that the
representatives of early It and Id could no longer be
kept apart, having in both instances got to be lit,
subject to be further modified by assimilation into
ll-ll, that is II, as in Welsh allawr, allor, ' an altar,'
Breton auter, Ir. altoir, from Latin altare ; call-
awr, ' a cauldron,' Bret, cauter, caoter, from Latin
caldarium — compare French chaudiere, ' a boiler or
copper ; ' cyllell, ' a knife,' from Lat. cultellus—
compare French couteau ; ellyn, ' a razor,' Bret.
aotenn, Ir. altan. In several of these words this
was an accomplished change in 0. Cornish; for
example, we have ellyn and cyllell in the later
Oxford Glosses written elinn and celleell, and still
earlier we find callawr written colour in the weU-
known 0. Welsh triplets beginning " Niguorcosam
nemheunaur " in the Cambridge Codex of Juvencus.
This proves that the Welsh had the sound which



(delwedd B60)


we write II as early as the 9th century, and
could pronounce it between vowels, as we do, a
point in which Welsh contrasts with Icelandic,
which also has the sound, but only before t. My
attention was called to its presence in that lan-
guage by an Icelandic gentleman in Oxford asking
me one day when such and such a college was
' ' buillt." On inquiry I found that this is the
sound which I always has before t in modern Ice-
landic : thus Icelandic kolt, ' a small forest,'
sounds to me like our kollt, ' a chink,' though
it may be that the Icelanders do not force the
breath so much to the right side of the mouth as
we do in pronouncing our II, which is sometimes
called unilateral by phonologists — it does not,
however, I may observe, deserve to be so called
any more than our I, which we pronounce also on
the right side of the mouth ; and so too, I suspect,
some Englishmen do. Look at these points as you
may, the coincidence between Welsh and Icelandic
is a striking proof that t has an affinity for U
which requires a physiological explanation.

Now we come to cases which do not involve
mutes, but only I, r, n: let us take first U and
Ir. The instances readiest to hand of II, that is
l-l yielding in Welsh the spirant surd which we
write II, occur in loan-words from Latin, such as
porchell ' a young pig ' from porcellus, ystafell ' an



(delwedd B60)


apartment' from stabellwm, Tstim/U 'Epiphany'
from Stella. Then there are other cases like Welsh
oil ' all,' Irish uile, from a stem olja, and Welsh
paiell ' a tent ' from Lat. papilio, which in Welsh
mouths became, no doubt, papiljo, that is to say,
if that was not the first and only pronunciation
which they heard from the Romans themselves.
But how did papiljo become pepyll, whence our
nxo^Qvn pabell ? did it become papilla with II for Ij,
ox papil'jo, papilljo, with I', II, produced by provec-
tion ? On the whole, I am inclined to take the
latter view as the more probable. Of Ir I have no
certain instances : so the next combinations are rr
and rl. As to the former, it makes in Mod. Welsh
rrh and rh, as for instance where a noun is preceded
by the definite article yr, 0. Welsh ir, which is a
proclitic. Take the following : y rhan ' the share '
for yr + ran; oW rhan ' from the share ' for o -\-yr
-\-ran; Vr rhan 'to the share' for i -\- yr -{ ran ;
and so in other cases, though rhan is regarded as
the radical form, of which more anon. The pro-
vected form of rl is written rll, as in perllan ' an
orchard,' oerllwm ' cold and bare,' garlleg from the
English garlic, and jarll from English eorl or
earl. But the importance of this change appears
mostly in the case of the definite article, as in y
llaw ' the hand ' for yr + llaw, d'r Haw ' from the
hand ' ion o-\-yr-\- llaw, a^r llaw ' with the hand '



(delwedd B60)


for a + yr + Ham ; and so on. Here it is to be re-
marked, as to the article prefixed to feminines,
that the parallels to y ddafad ' the sheep ' for
yr + da/ad, y formyn ' the maid ' for yr + morwyn,
are to be sought not in y Haw ' the hand ' for
yr + Haw, and y rhan ' the share ' for yr + rhan,
but in an earlier stage yr + law and yr + ran, which
passed into y(r) Haw and y{r) rhan. There still
remain to be noticed nl and nr, the provected
forms of which are written nil and nrk as in gwinllan,
' a vineyard ' and enllyn anything eaten or drunk
with bread, such as butter, cheese, milk, beer, or
the like : so also after the preposition yn, as in
yn Llundain ' in London ' and yn llawn ' in full.'
Whether and in what cases I has passed immedi-
ately into ll and not through an intermediate Ih,
which would be the parallel to rh, I am unable to
decide. But both ll and Ih would be provected
forms of I, and we seem to detect a trace of the
latter in 0. Cornish in the later Oxford Glosses,
which give us the equivalent of our enllyn, Ir.
anion, in the form ehnlinn, whereby is probably
meant enlhinn or e\nr]hinn.

A word now as to ll and rh initial. Ll and rh,
whether initial or not, are confined, as far as con-
cerns the Celtic languages, to Welsh and Cornish
— Edward Llwyd found traces of both in Cornish.
But the fact that they are foreign to the Breton



(delwedd B60)


dialects seems to show that they date after the
mutual diiferentiation of Welsh and Breton. We
have no Welsh manuscript authority for rh in
the 0. Welsh period, but II is found written in the
Black Book of Carmarthen of the 12th century as it
is now. On the other hand, 0. Cornish offers an
instance in the later Oxford Glosses of a word
beginning with hi identical probably with Ih : it is
hloimol, which is unfortunately as obscure as the
Latin glomerarium which it was intended to ex-
plain, but the Mod. Welsh equivalent might be
expected, if it existed, to begin with the syllable
llwyf. But how, you will ask, is the provection of
initial I into Ih, II, and of initial r into rh, to be
accounted for ? The first answer to suggest itself
is, that it is the result of the influence of the other
consonants, which as initials remain c, t, p, &c.,
while as medials or quasi-medials they are reduced
to g, d, b, &c. Thus .c initial and g medial would
be matched by II initial and I medial ; and so
with rh and r. Supposing that it could be shown,
but it is hardly probable that it can, that the. pair-
ing of U and I, rh and r, began some time posterior
to that of c and g, t a.Vid d, and so on, this might
be admitted as a passable explanation, though it
would be open to the objection that the analogy
of c, g, for insta'nce, would require I and r as
initials to remain unchanged, but to give way as



(delwedd B60)


medials to some softer sounds, l^ and r^ ; and this
applies both to Welsh and Sassarese, the agreement
between which extends to r. Thus in Welsh we
say rlmyd ' a net,', but ei rmyd ' his net,' and the
Sassarese word for net is pronounced rrezza, while
the net is, nevertheless, la rezza. But one could
not, in the way here suggested, account for initial
r always appearing in Ancient Greek as p, a
coincidence with Welsh which can hardly be acci-
dental ; nor is this all, for in Ancient Greek, as in
Welsh, two r's coming together resulted in pp as in
nvppo<!, KdXXippdr}, which the Komans transcribed
Pyrrhus, Calirrhoe — the distinction between p and p
is unknown in Mod. Greek. On the whole, then,
nothing remains but that we should ascribe the
distinction between the liquids as initials and non-
initials to the same cause, to a certain extent, as
that between the mutes. Thus from the facts of
mutation already discussed, as, for instance, of c
becoming g when non-initial and following a vowel,
while initial c undergoes no such a change, it seems
to follow that initial c, owing wholly or in part to
its position, is pronounced with more force than
when it happens to be preceded by a vowel. The
same applies to other mutes, and herein Italian, as
has already been mentioned, is at one with the
Celtic languages. Moreover, the greater force of
initial consonants has been established by direct



(delwedd B60)


measurement in a way which must now be briefly

In an address to the Philological Society, Mr.
Ellis gave a short account of an. instrument called
the logograph, invented by Mr. W. H. Barlow for
recording, among other things, the comparative
force of articulation in speech. Since then Mr.
Barlow has very kindly answered various' queries
I have sent him, and favoured me with a copy
of his own description of his invention to the
Royal Society in a paper entitled : " On the
Pneumatic Action which accompanies the Articu-
lation of Sounds by the Human Voice, as exhibited
by a Recording Instrument. By W. H. Barlow,
F.R.S., V.P. Inst. C.E." (Proc. of the Roy.
Soc, vol. xxii. pp. 277-286). " The instrument I
have constructed," he says, " consists of a small
speaking-trumpet about four inches long, having
an ordinary mouthpiece connected to a tube half
an inch in diameter, the other end of which is
widened out so as to form an aperture of 2 J inches
in diameter. This aperture is covered with a
membrane of goldbeater's skin or thin gutta
percha. A spring which carries the marker is
made to press against the membrane with a slight
initial pressure, to prevent as far as practicable
the effects of jar and consequent vibratory action.
A very light arm of aluminium is connected with



(delwedd B60)

ńLECTURE ir. 81

the spring and holds the marker ; and a continu-
ous strip of paper is made to pass under the
marker in the same manner as that employed in
telegraphy. The marker consists of a small fine
sahle brush placed in a light tube of glass one-tenth
of an inch in diameter. The tube is rounded at
the lower end, and pierced with a hole about one-
twentieth of an inch in diameter. Through this hole
the tip of the brush is made to project, and it is fed
by colour put into the glass tube in which it is held.
To provide for the escape of the air passing through
the instrument, a small orifice is made in the side
of the tube o'f the speaking-trumpet, so that the
pressure exerted on the membrane and its spring
is that due to the difference arising from the quan-
tity of air forced into the trumpet, and that which
can be delivered through the orifice in a given
time." The line described by the marker when
the instrument is used looks somewhat like the
outline of a series of valleys and mountains repre-
sented in section : the valleys are the vowels, and
the high pointed peaks the surd mutes c, t, p,
while the other oral consonants are represented by
lesser and less sudden elevations. Among the re-
sults of Mr. Barlow's experiments on the logograph
may be mentioned the following : —

The pneumatic force of the vowels is compara-
tively small.



(delwedd B60)


That of sonant consonants is greater, but falls
considerably short of that of the surd mutes c, t, p.

That of an initial consonant is greater than
that of the same consonant preceded by a vowel.
This, in answer to one of my queries, has been as-
certained by Mr. Barlow, who has very kindly sent *
me the diagrams in respect of c, t, p, g, d, b.
Thus it would seem that the greater force of an
initial consonant depends on a physiological cause,
and that it is its continued influence on the pro-
nunciation of initial I and r which brought about
their provection into II and rh respectively.

Assuming, as I think we now may, this initial
pressure to be a vera causa, we can apply it to
explain another feature of Welsh phonology. I
allude to our gm for m semi- vowel ; for as the Ital-
ians derive their guaina, ' a scabbard,' from Latin
vagina, and the French their guerre from a word
the form of which recalls its English equivalent
war, so Welsh regularly makes use of gw, formerly
written gu,, for Aryan w, which it is the custom
of glottologists to treat as v. Thus Latin vinum
becomes in Welsh gwin, ' wine,' and the same rule

* It was only lately that it occurred to me to ask Mr. Barlow to
experiment on initial I and r, and as he waa on the point of setting
out for Philadelphia, and the instrument had been lodged in the
Kensington Museum, I am unable to give the results of a direct
experiment on I and r. However, I have no doubt that they fol-
low suit with the other consonants mentioned.



(delwedd B60)


is followed in native words such as gwynt, Latin
ventus, Eng. wind. In Old Welsh this was not
confined to the beginning of a word — witness pet-
guar, now pedwar, ' four ; ' but, as in the case of
pedmar, the g disappeared later. However, initial
gw is not in sole possession, as it is occasionally
supplanted by ckw. Thus cAmertkin, ' to laugh,'
and ckwareu, ' to play,' have, as far as concerns
Mod. Welsh, driven gwerthin and gwareu out of
the field ; while chwannen, ' a flea,' is the only
form, gwannen being altogether unknown, though
the word is probably of the same origin as the
German wanze, ' a bug.' To these may be added
a remarkable instance in the case of a Latin loan-
word : vesica becomes in Welsh either chTvysigen
or gwysigen, ' a bladder, a blister.' Looking at
these facts — initial gw, initial chw, and w for
medial gw — the common combination from which
we must set out, can hardly but be assumed to have
been^^w, with gh pronounced as a very soft spirant
like the g one sometimes hears in German sagen,
' to say.' In Old Welsh this combination would
of course be written gu ; but where it occurred in
the body of a word, the guttural would eventually
drop out of the pronunciation, whereas, occurring
initially, it would come under the pneumatic pres-
sure which has just been supposed, to have induced
the provection of / and r into II and rh ; and the



(delwedd B60)


result would be the like provection (A gh into eh.
That of gh into g differs from them in its resulting
in a mute and not a spirant, but it may be com-
pared with the Sassarese substitution of a strong h
for an initial v as in bozi, Ital. voce, while as a
quasi-medial in la bozi, ' the voice/ the labial has
the weaker pronunciation of a kind of v or Spanish
b. As for the transition from w to gkm, it can hardly
have taken place all at once : it happened, pro-
bably, through the intermediate stage of 'w, where
the soft palate was just slightly moved by the air
in its passage from the larynx into the mouth
during the pronunciation of the w. But why the
soft palate should have been drawn in at all is an-
other of those questions which I must leave to the
student of the physiology of sounds. It is to be
noticed that the guttural preceding the semi-
vowel dates from the 7th or the 8th century, as no
trace of it is to be found on our early inscribed
stones, which show only F, or, in Ogam, a charac-
ter which is to be read w.

In the case of U and rk, the difference between
Welsh and Irish was owing to a change on the
part of Welsh only : in the present instance the
gulf has been widened by changes on both Welsh
and Irish ground. The former have just been
described, and the latter consist in dropping the
semi-vowel, as a rule, where we have reduced 0.



(delwedd B60)

ńLECTURE ir. 85

Welsh gm into w, while, as an initial, it was some
time or other modified from w to v, which was sub-
sequently provected into f, for seemingly the same
reason that gh, I, and r initial became in Welsh ch,
II, and rh respectively. All this happened before the
date of the earliest Irish manuscripts of the 8th
century, but no trace of it is known on the Ogam-
' inscribed stones of Ireland : on them the semi-
vowel is represented by the same character which
I would read m on British monuments. The time
may be still more narrowly defined: the change
had not taken place before the middle of the 5th
century, as seems to be indicated by the fact, that
an Irish saint, supposed to have died about 460,
bore a name which in Ireland afterwards became
Fingar, and in Cornwall, where he spent a parb of
his life, Gwinear, as it is now written. This
implies that in his time his name did not com-
mence with an f, but with nearly the same initial
in Ireland and Cornwall, namely w or v. More-
over, about the beginning of the 6th century the
semi-vowel was still pronounced in Irish where it
has since been elided. Thus in one of the lives of
the Irish saint Monenna or Modvenna, a contemporary
of St. Patrick, she is spoken of as a virgo de
Convalleorum populo, another gives the last words
as Conalleorum populo, and a third makes her a
native of terra Conallea, which must, I suppose, be



(delwedd B60)


Tyrconnell. Now Conall is one of those vocables
which have dropped the semi-vowel, which is excep-
tionally retained in the Convalleorum alluded to :
the "Welsh is Cynmal, 0. Welsh Congual, and still
earlier ovnovali on an inscribed stone in Cornwall.
It was thought right to dwell on Welsh ^w = Irish
f at some length, not only on account of their
phonological interest, but because they are not
infrequently relied upon as evidence of a very
profound and primeval difference of language
between the Irish and the Welsh.

Now that we have fairly come to the end of our
task — at least in outline — as far as regards the
consonants, than which we have no reason to
suspect the vowels of being less interesting, though,
maybe, the laws they obey are more subtle, we
may be allowed to indulge in a few remarks of a
more general nature. Enough has probably been
said to convince you that, in spite of our having
reserved to the last the fag-ends of the subject,
Welsh phonology is far from devoid of interest.
The regularity which pervades it leaves but little
to be desired, and it falls, comparatively speaking,
not so very far short of the requirements of an exact
science. Some, however, have no patience with a
discussion which turns on consonants and vowels,
and nothing short of etymologies bearing directly
on ethnological questions or the origin of lano-uafe



(delwedd B60)


can hope to meet with their approval. This need
not surprise any one, for, as a rule, few people
feel interested in the details of a scientific inquiry,
and duly realise the fact, that what they regard as
food only fit for the shrunken mind of a specialist
must necessarily precede those gushing results they
thirst after. In the case before us, we are only too
familiar with the worthlessness of the fruits of a
method which ignores the phonological laws of the
language with which it pretends to deal, or fails
to do justice to their historical import ; and it is
by his attitude with respect to these laws that one
can generally tell a dilettante from a bona fide
student of the Celtic languages. The former, you
hardly need be told, never discerns a difficulty ;
for to him a letter more or less makes no difference,
as his notion of euphony is so Protean that it is
equal to any emergency ; but the latter frequently
stumbles or goes astray, and has to retrace his
steps ; and altogether his progress can be but slow :
so much so, in fact, that some of the leading
glottologists of our day think it on the whole
impossible to attain to the same state of knowledge
respecting the history and etymology of Celtic
words as that arrived at in the case of the other
Aryan tongues. That it is harder is certain, but
that it is impossibk I am inclined to doubt. At
any rate, so far progress is being made ; nor is there



(delwedd B60)


anything which may be regarded as an indication
that we have nearly come to the end of our tether.
For example, one of the tasks — and only one out
of several — which the student of an Aryan language
proposes to himself is to discover, as far as that is
practicable, the origin of every word in its voca-
bulary, and to show to what recognised group of
words it btelongs, or, in other words, from what
root it is derived and how. In some of the lan-
guages kindred to our own this work has already
been carried very far, and the nunjber of the
vocables in them of obscure origin has been
materially reduced; but in the Celtic languages
this search, being attended with greater difficulties,
is not so far advanced ; but it is going on and
likely to go on, as you will see on perusing the
Eevue Celtique or Kuhn's Beitraege, where you will
find, among others, some of the most stubborn
words of our vernacular forced, one after another,
to surrender the secrets of their pedigree.

But whence, it will be asked, does this greater
difficulty attending the study of the Celtic lan-
guages, and of Welsh in particular, proceed?
Mainly from two causes — the great dearth of speci-
mens of them in their earlier stages, and the large
scale on which phonetic decay has taken place in
them. For, to pass by the former for the present,
it is to be remembered that the phonetic changes



(delwedd B60)

ńLECTURE ir. »y

which have been engaging our attention are but
the footprints of phonetic decay, and that the
phonological laws which have just been discussed
form but the map of its encroachments and a
plan, as it were, of its line of attack. With
these before our eyes, we are, to a certain extent,
enabled to infer and picture to ourselves the posi-
tions, so to say, and the array in which the forces
of our language were at one time drawn up. So,
when you hear it said, as you frequently may, that
Welsh or Irish is the key to I know not how many
other languages, do not believe a word of it : the
reverse would be nearer the truth. We want con-
centrated upon the former all the light that can
possibly be derived from the other Aryan tongues ;
that is to say, if we are to continue to decipher
their weather-worn history.

( 90 )



(delwedd B60)


" La dissonanza tra lingua e lingua, se pur non sia minors, riesoe di
oerto, in generals, men sensibile rispetto aUe vooali ohe non rispetto
alle consonant! ; ma appunto per questo, torna piil logioo, in una trat-
tazione come la nostra, che il ragguaglio delle cousouanti sia mandato
innanzi a quello delle vocali." — G. I. ASOOLI.

At first it was not my intention to notice the
vowels, but it has since occurred to me, that if
they were to be passed over in silence, you might
suppose that I endorse the first part of Voltaire's
definition of etymology as a science in which the
vowels are of no consequence and the consonants
of extremely little. But there is another reason
why they should be noticed here, and that is the
fact that without taking them into account the his-
tory of the consonants cannot be thoroughly under-
stood. Before, however, proceeding to any details
it will be necessary roughly to indicate what
vowels in Modern Welsh represent the vowels of
the Aryan parent-speech respectively. It is to be
observed, that, as a result of the researches of
Professor Curtius of Leipsic, and others,- it is now
generally accepted as a fact that the Western
Aryans not only retained the vowel a in some



(delwedd B60)


words, but also changed it into e in others ; but it
would make no difference, so far as our present sub-
ject is concerned, if it should some day be made
out that the parent-speech had two or more kinds
of a (as is the case, say, in English), which the
Eastern Aryans confounded in course of time, and
reduced to one, while their brethren in the West
never completely effaced the distinction between
them. It further appears probable that, anterior
to the separate existence of Irish and Welsh, a
had also been modified in not a few words into o
in the common Celtic from which these languages
have branched off. Thus while Sanskrit harps on
the same string of a, the Celtic and other Aryan
languages of Europe have no less than three vowels
at their disposal, namely, a, e, o : witness our tad
' father,' de^ ' ten,' and pob-i ' to bake,' which are
in Sanskrit respectively tata, dagan, and pac, all
with a. So far, then, as concerns Welsh or Irish,
we may treat the following vowel-sounds as ori-
ginal : a, a, e, e (?), i, i (?), o, u, u (?), ai, au. Our
task is now briefly to point out the most common
and direct continuators of each of them in our lan-

A. The a of the Aryan parent-speech is retained
in the following words and many more which
might be enumerated : —



(delwedd B60)


aden, ' a wing,' adar, ' birds/ from pat, whence also
Skr. pat, ' to fly,' Gr. irrepov, 'Eug. feather.

a/al, ' an apple,' Lith. obolys, 0. Bulg. jahluko,
Eng. apple.

am, ' about, around,' 0. Ir. imb-, imm, Grr. afi^l,
Lat. amb- in ambages, Ger. um.

an-, as in annoeth, ' unwise,' anamserol, ' untimely,'
Ir. an-, Skr. an-, Gr. dv-, Lat. in-, Eng. un-.

ar, ' ploughland,' arddu, ' to plough,' aradr, ' a
plough,' Ir. arathar, ' a plough,' Gr. apoto, * I
plough,' Lat. aro, same, ardtrum, ' a plough,'
Goth, arjan, ' to plough,' Eng. to ear, earth.

all-, in alltud, ' one of another nation,' Ir. aile, Gr.
aWo?, Lat, alius, Eng. e/se.

arcA, ' a bidding, a request,' from ?aek, whence also
Latin precor, ' I pray,' Ger. /rage, ' a ques-
tion : ' another form of the same root seems
to be PAKSK, whence Skr. prach, ' to demand,
to ask,' Lat. posco (=porsco), Ger. forschen,
' to inquire, to investigate.'

cad, 0. Welsh cat, ' battle, war,' whence Catteyrn,
'battle-king,' Early Welsh Catotigirni; Ir.
cath, Gaulish catu in Caturiges, Catuslogi ;
Early Eng. heatho-, ' war, battle.'

caled, ' hard,' Zend gareta, ' cold,' Eng. cold,
which seem to show that the common base
was scareta, and that the Celts reasoned from
cold to solidity.



(delwedd B60)


can, ' a song,' canu, ' to sing, to crow,' Ger. hahn,
' a cock.'

i, ' seed/ Lat. satus.
ha/, 'summer,' Skr. samd, 'year,' Zend hama,

' summer.'
halen, ' salt,' kallt, ' salty, salted,' Ir. salann,

' salt,' Gr. a\?, Lat. scd, Eng. salt,
pa, ' what,' Ir. ca, Skr. kas, ' who,' Lat. quo-, in

quod, Goth. Avas, Eng. who.
pas, ' the whooping-cough,' Skr. kds, ' to cough,' 0.

Eng. hoostan, ' to host, to cough,' Ger. Austen,
tarw, ' a bull,' Ir. tarbk, Gaulish tarvos, Zend.

thaurva { = tharva), 'violent, strong, hard,'

Lat, tormts : it is not certain that these words

are connected, but in any case tarw cannot be

identified with the Latin taurus.

E. The vowel e for Aryan a occurs in Welsh, in
common with other European languages, in a
good many words, of which the following are
a few : —

ad-fer, ' to restore,' from the same origin as Gr.
^e'joo), lidl.fero : Skr. bkar, ' to 6ear.^

cred, 'belief,' Ir. creitem, Lat, credo : Skr. graddhd,
'trusting, faithful.'

chwech, ' six,' Ir, s4, Gr. e^, Lat. sex : Skr. skask.

deg, ' ten,' Gr. Ima, Lat. decern, Goth, taikin, Eng.
t&a: Skr. daqan.



(delwedd B60)


deheu, ' right, south,' 0. Ir. dess, Gr. le.^io'i, Lat.

dexter: Skr. dakshina.
ebol, ' a colt,' 0. Ir. ech, ' a horse,' Lat. equus, 0.

Eng. eok: Skr. agva.
gen, ' the chin,' Gr. yew;, Lat. getia, Goth, kinnus,

Eng. chin: Skr: hanus.
keb, 'besides, without,' 0. Ir. seek, Lat. secus.
heb, ' quoth,' Gr. taireve. (theme aetr), Lat. in-sece,

Lith. sakau, ' I say,' Ger. sagen, Eng. say.
^ew, ' old,' 0. Ir. sen, Gr, ei/i; («at z/ea), Lat. senex :

Skr. sa?2a.
met, ' honey,' Ir. mjY, Gr. fjJXi, Lat. jwe^, Goth.

mellt, ' lightenings,' 0. Prussian, mealde, ' a

lightening,' 0. Bulg. mliinij.
melyn, ' yellow, tawny,' Gr. fiekav, gen. fjdkavo'i,

' black, dark, blue,' Lith. mUynas, ' blue : '

Skr. malina, ' dark, black.'
merck, ' a girl, a daughter,' Lith. merga, ' a girl.'
nef, ' sky, heaven,' 0. Ir. 7wm, Gr. ve'^o?, Lat.

nebula, 0. Bulg. nebo, ' heaven.'
ser, ' stars,' Gr. da-rrip, Lat. Stella, Eng. star : Skr.

staras, ' stars.'
serch, ' love, affection,' Ir. searc, Gr. arepya, ' I

love,' a-Topryrj, ' love or natural affection.'

7. Aryan t is represented in 0. Welsh by i,
written y in Mod. "Welsh, and i or y indiffe-



(delwedd B60)


rently in the intervening period. But in
most cases the y of Mod. "Welsh has taken
the place of other vowels, while the instances
where it is the representative of an i of Aryan
or even European standing are comparatively
few. The following may be mentioned : —

dyw, ' a day,' he-ddyw, ' to-day,' Gr. evSto? ( =
evBifo^), ' at midday,' Lat. diu, diurnus
(= dius-nus) : Skr, diva, ' heaven, day.'

hysp, fem. kesp, ' dry, not giving milk,' Gr. layyo^,
' dry,' Lat. siccus, Zend hisku, ' dry ; ' the
"Welsh, the Greek, and the Zend forms seem
to be the results of reduplication — si-siqv- or

mysc, as inyn mysc, ' in the midst of,' Gr. fxiyvvfu,
/ttffyo), Lat. misceo, Eng. mix, Skr. miksh.

nyfio, ' to snow,' from a root snighv, whence also
Gr. vi^ei, Lat. ninguit, ningit, or nivit, ' it
snows,' Eng. snoK : Zend ^izA.

py, ' what, which ' (now superseded by pa), Gr.
Tt?, Ti, Lat. quis, quid, Oscan pid: Skr. kirn.

yd, ' corn,' 0. Ir. ith, gen. etho, Lith. pttus,
' mid-day, mid-day meal : ' Skr. pitu, ' food,

0. In a good many instances o has taken the place
of a, at a date probably falling within the
limits of the history of the "Welsh language,
but in others it seems to be, as already



(delwedd B60)


suggested, of older standing, as may be
gathered from its appearance in the corre-
sponding forms in other languages nearly
related to Welsh, as in the following in-
stances : —

coll, ' hazel,' 0. Ir. coldde, ' columns,' Lat.
corulus (= cosulus), Eng. hazel.

dqf, ' tame,' Lat. domare, ' to tame,' Eng. tame.

mock, ' soon, quick,' Ir. moch, Lat. mox; Skr.

iweth, 'naked,' Ir. nocAt, Lat. nudus {^:=no{g)vidus),
Goth, naqvaths, Eng. naked.

nos, ' night,' kenoetk, ' to-night,' trannoetk, ' over-
night, the day after,' literally trans noctem,
Mod. Ir. anockt, ' to-night,' Gr. vv%, gen.
vvKTo^, Lat. now, gen. noctis, Lith. naktis,
Goth, nakts, Eng. nigkt: Vedic Skr. nakti.

( = op), as in o thry efe, ' if he turn,' Lat. nec-
opinus, in-opinus, opinio, 0. Norse, ef, if,
' doubt,' Ger. obi Eng. if.

og or aged, ' a harrow,' Lat. occa, Lith. akeczos,
ek^czos, 0. H. Ger. egidd. Mod. Ger. egge.

pobi, 'to bake,' Gr. iriaam, future Tre'^o), Lat.
coquo : Skr. pac.

m/tk (for oitA = oct), ' eight,' Ir. ocAt, Gr. oktoi,
Lat. octo, Eng. eigkt: Skr. ashman.

U. Aryan m is represented in 0. Welsh by u,
written in Mod. Welsh w: however, the in-



(delwedd B60)


stances where the original u may perhaps not

have been modified are comparatively few,

such as the following : —
cnm, ' dogs,' Ir. con, Gr. «we9, Lat. canes, Bng.

hounds. Ski. ^dnas, gunas.
drwg, ' bad,' Ir. droch-, Ger. trug, ' deception,' he-

triigen, 'to deceive,' Skr. druh, 'to injure,, to

harm,' Zend druj, ' to lie.'
dwfn, '.deep,' 0. Ir- domnu, ' depth,' Lith. dubus,

' deep, hollow,' 0. Bulg. duno (= dubno),

' ground ' (compare Ir. domhan, ' the world '),

Goth, diups, ' deep,' 0. Eng. deop, Mod.

Eng. deep.
}ud, in the 0. Welsh names Judgual, Margetjud,

now Idwal and Meredudd, comes from the root

TUDH, whence also Ir. iodhnach, ' armed,' Gr.

vafilvTi, ' a battle,' Skr. gudk, ' to fight,'
rhwd, ' rust,' Lat. russus, Ger. rosten, ' to rust,'

Eng. rust, from the root etjdh, whence Welsh

rhudd, ' red,' and its congeners.

A. Aryan d seems to have in Early Welsh
acquired a guttural sound, which passed into
6, yielding in Mod. Welsh o and aw, the latter
being used in monosyllables, and the former
in most other words as Welsh is now pro-
nounced ; the instances are numerous — take
the following : —



(delwedd B60)


brawd, ' a brother,' pi. brodyr, Ir. brdthair, Lat.

frdter, Eng. brother, Skr. bhrdtar.
chwiorydd, ' sisters ' (sing, chmaer), Lat. sorores,

Eng. sisters, Skr. svasdras.
dawn, ' a gift in the sense of talent or genius,' Ir.

dan, Lat. ddnum, 0. Bulg. dam.
llawn, ' full,' Ir. Ian, Lat. plenus, Skr. prdna.
llawr, 'floor,' Ir. Idr, 'Eiug.Jloor.
modryb, ' an aunt,' from the. word for mother,

which is lost in Welsh, but is in Irish mdthair,

Gr. fj.^TTjp, Doric fidrrip, Lat. mdter, Eng.

mother., Skr. mdtar.

E, 1. It is not supposed that the parent-speech
had e, and it is doubtful whether it had i':
even supposing that it had the latter, I have
failed to trace a single instance down to Welsh.
The nearest approach to this would be the
case of Welsh byw, ' quick, living,' 0. Ir. beo,
bin, and Welsh byw, ' a life or lifetime,' 0. Ir.
biu, in Fiacc's Hymn (Stokes' Goidelica, p.
128), Greek yS/o? ; but Latin vlvus, Sanskrit
jiva, and their cognates can hardly be said to
prove beyond doubt that the i was originally
long. It is, however, probable that e had
replaced d in & few Celtic words, or even
passed into i, before the separate history of
Welsh or Irish can be said to have begun. The



(delwedd B60)


instances alluded to are those where Welsh and

Irish have i answering to Latin e, as follows : — '
gwir, ' true,' Ir. /ir, Lat. verus, Goth. vSrjan (in

tuzverjan), ' to believe,' Ger. waAr, ' true.'
hir, ' long,' Ir. sir, Lat. serus, ' late,' Goth, seithus,

rhi, ' a king,' 0. Ir. ri, gen. rig, Gaulish Dumno-

rix, Dubno-reix, Dubno-rex, Catu-riges, Lat.

rex, gen. regis; Goth, reiks, Skr. rdjan.
tir, 'land,' Ir. tir, Lat. terra, ' the earth.'

U. Nearly the same remark applies to u as to i.

Ai. From the different representatives of at in
the various Aryan languages it has been
inferred that the. primitive Aryans had two
kinds of this diphthong, which glottologists
would distinguish as ai and di: the case
is, however, not quite so clear as it looks in
some books. Now, at a certain stage in the
history of Welsh, ai had become oi, which
has since been differentiated by causes to be
noticed later into oe and m/ in Mod. Welsh.
The ordinary Irish representatives are ia and ^.
The following words are instances in poiiit : —

bloesc, ' imperfect or indistinct in one's pronun-
ciation,' Skr. mleccha, ' a foreigner, a bar-
barian :' Sanskrit ch = sk.




coed^ 'wood, trees/ Ir. ciad-cholum, 'a wood-
pigeon,' Lat. bu-cetum, ' a pasture for cattle,'
Goth, haitki, ' a heath, field,' haithivisks,
' wild,' Eng. heath, h,eathen.

coel, ' angury, superstition, belief,' Ir. eel, Goth.
hails, 'whole, uninjured,' hailjan, 'to cure,'
Eng. heal, health.

drcyf-, in dovyfol (also dwymol), ' divine,' 0. Ir. dia,
gen. dii, ' God,' Lat. divus, Skr. deva, ' god-
like, divine, a god.'

hmy (= sa-i), hroynt, ' they, them,' Ir. iad, Gr. ot, ai.

pToy, ' who,' Ir. cm, cS, Lat. ^Me?, g;Mae (more com-
monly qui, qu(B), Umbr. poi, ' who ' — the
same particle i appears for instance in the
Lat. hcec (= ha-i-oe), and Gr. ovrocri.

Ai. Aryan di makes u in Welsh, now pronounced
nearly like the u of the Germans. It was
derived from di by a process similar to that
whereby ov assumed the sound of v in Modern
Greek, before both became identical with I in
pronunciation. The Old Irish equivalent was
oi or oe, now written ao (aai), and pronounced
in some parts like the uee of queen accord-
ing to O'Donovan : as pronounced in Galway,
it seems to me to lie between our Welsh u
and i. The following instances may here be
mentioned : —



(delwedd B6101)


cut, ' narrow,' Ir. caol.

cynud, ' fuel,' 0. Bulg. gnetiti, ' to kindle,' 0.
Prussian, knaistis, ' a firebrand,' 0. H. Ger.
gneisto, ' a spark.'

hud, ' a charm, a spell,' Lith. saitas, ' sorcery,' 0.
Norse seidhr, ' a kind of sorcery or magic,'
Ger. seid.

hiifen, ' cream,' 0. H. Ger. seim, Mod. H. Ger.
konig-seim, ' run-honey,' Eng. seam, ' lard,'
whence our saim, ' grease,' has heen borrowed.

tu (for ttcf), ' side,' Ir. taobk.

ud-, in anudon, ' a false oath, perjury,' 0. Ir. oetk,
Goth, aiths, Ger. eid, Eng. oath.

un, ' one,' 0. Ir. oin, Mod. Ir. aon, Lat. oinos
(later unus), Goth, ains, 0. Eng. an, Mod.
Eng. one, atone, only, an — the pronunciation
of one as nun was originally that of a parti-
cular dialect like routs for oats, and an is the
Old Eng. an (that is an) shortened owing to
the proclitic pronunciation of the numeral
when used as an indefinite article : the Ger-
mans of late sometimes distinguish an and
one as ein and iin respectively.

Au. Even supposing that the primitive Aryans
distinguished two kinds of au, which is ex-
ceedingly doubtful, it seems to be quite
hopeless to separate their respective repre-



(delwedd B6102)


sentatives in the modern languages of the
Celts. In Welsh they are u and uw (pro-
nounced like German il followed by German
u) ; the latter is used only in a few words,
mostly before ch; otherwise u and uw take
their places like o and aw. The Irish equi-
valents are ua and 6. Take the following
instances : —

dun, ' a knee/ Lat. clunis, Lith. szlauniSy Skr.

rhudd, ' red,' Ir. ruadh, Lat. rUcfus, Goth, rauds,
Ger. roth, Eng. red.

tud, ' nation, country,' Breton tud, ' men, a people,'
Ir. tuatk, ' a people, a nation,' Gaulish toutius,
Oscan touto, Goth, thiuda, Ger. Deutsch,
' Dutch or German.'

buwch, ' a cow,' pi. buchod, Cornish biueh, Breton
bioc'k, all with a final s irregularly represented
by ch, but bu and buw also occur in Welsh, Ir.
bo, Gr. j8ov?, Lat. bos, Eng. cow, Skr. nom.
gaus, gen. gos.

Duw, ' God,' also Duwch with ch (as in buwch),
and only vulgarly used in Duwch anrcyl!
which corresponds to the German exclamation
Hu lieber Gott! Gr. Zeu?, voc. Zeu, Lat. Jou^
piter, Skr. nom. dyaus, voc. dyaus, ' sky,
heaven,' Byaushpitar, ' Heaven-father.'

Mw«?, 'porridge,' 0. Cornish iot, Breton iot, 0.



(delwedd B6103)


Ir. ith, Lat. jAs, ' broth,' Lettish j&ut, ' to
mix meal up in \Ater,' Skr. yws, yusha, 'pea-
soup,' d-yavana, ' axpot-ladle or some similar

uchel, ' high,' uwck, ' higher,' uckqf, * highest,' Ir.
uasal, * high, noble,' Gaulish uxel-, in Uxela,
Uxellodunum ; and probably ov^a/xa in Pto-
lemy's Ov^afia BapKa is identical with our
uckaf, so that we might call the place ' Upper
Barca : ' the root would seem to have been
auks (as in Gr. av^dva) from aug, as in Lat.
augeo, ' I increase,' auctus, ' enlarged, in-
creased, great, abundant,' 0. Prussian auktai-,
' high,' Lith. auksztas, ' high.'

Cnunch, cuwch, lluwch, rhuwch, are other Welsh
words with uw, which is replaced by u when a
syllable is added, but their origin is obscure.

The foregoing are a few points which it was
thought necessary to mention in the vowel system
of Welsh : now some of the principal changes and
modifications which have obtained in it must be
considered somewhat more at leisure. Some of
them, such as those involved in the history of aw,
wy, uw, have already been touched upon. For it
is impossible, language being in -a constant state
of flux and change, to discuss its organism alto-
gether apart from its pathology, so to say, however



(delwedd B6104)


neat such a plan may look in theory. To begin
with the evolution of aw from d, this seems to
mean that d passed in the course of time into a
sound identical, or nearly identical, with the
English Yowel in hall and draw, and that, where
it was not eventually shortened, yielding o, it was
diphthongised into au, which we now write aw.
As to the date of the transition, no instance of
au occurs in the earlier class of Welsh inscrip-
tions, so it may be presumed that it did not take
place before the 7th century. For a parallel to it
we need not go further than English : take, for
instance, the Old English word stdn, that is stan,
which is now written stone, and pronounced stown
with a long o followed by a more or less percep-
tible w, or with some modification of that diphthong,
seldom if ever with a long o pure and simple. To
this might be added plenty more, such as bone,
home, rope, for the 0. English ban, ham, rap, re-
spectively. But for a perfect parallel consult
the Swabian pronunciation of German — witness
Schrcaub and aubend for Schwab and abend: nor is
the change unknown in Sanskrit.

With respect to oe and wy, it is not quite cer-
tain what the Kymric starting-point should be
assumed to have been. But reasoning backwards
from the loan-words which have wy in Mod.
Welsh for Latin e, one is led to the conclusion
that for some time after the Eoman occupation



(delwedd B6105)


the antecedent of my in native words naust have
also been e, or some such a diphthong as ei, which
could be taken for ^. Either & or ei would here
do, but the advantage of simplicity is on the side
of the former when one comes to assign the com-
mon Goidelo-Kymric prototype of Welsh my^ oe,
on the one hand, and Irish 4, ia, on the other.
So among the steps whereby d yielded oi, whence
rm/ and oe were differentiated, we should have to
reckon ei, ei, ai, which would make the series e,
ei, ai, oi. The earlier of these steps are fairly
exemplified in the ordinary English pronunciation
of such words as name, paper, as ne^m, pe'per,
neim, peiper, or even ndim, pdiper, with a long e
or a followed by a more or less marked i, which
so frequently mars the English pronunciation of
French words containing a long e, as the w
sound in stone does in that of French words
involving long o. The later steps in the series
are well known in Irish, where such instances as
croinn for crainn, genitive of crann, ' a tree,' boill
for baill, ' members,' and toibre, taibre, ' give,'
frequently occur, and illustrate a tendency which
is perpetuated in the Anglo -Irish pronunciation,
which makes the English words firie, I, line, into
foine, oi, loin, approximately.

In the case of u and urn, it is probable that the
Aryan au which they represent had become a



(delwedd B6106)


Goidelo-Kymric o (or ou), whence the Irish de-
rived their 6, ua, while the Welsh changed it into
a broad u, and later into the narrow u of Mod.
"Welsh. For this is the ordinary representative
of both Latin o and u, as in Uqfur, ' labour,' from
Latin labor-is, ffuvien, ' a line, a cord,' irovn funis,
and addurn, ' an ornament,' from adorn-o. In the
few native words already noticed this u was diph-
thongised into uw, and that, it would seem, at no
recent date, as we appear to detect traces of it in
the Breton bioc'k, ' a cow,' and the Cornish tot,
' porridge,' where the Welsh is bumch and uwd.

Before leaving these points, a word may not be
out of place as to the Irish ia and ua, or ia and
ua, as they are more commonly written : the i and
u are long, and followed by only a very slight
touch of a. They remind one somewhat of the
Lithuanian diphthongs ie and uo, also written e and
u. But whether the way they were arrived at was
the same, or nearly the same, is not evident : in
the case of the Irish ones the steps probably were
e, ^a, ia, and 6, da, ua, respectively. No certain
traces of either diphthong are known in the early
Ogmic inscriptions of Ireland, and they date, pro-
bably, after the 6th century.

Here it may be asked why such cases of vowel
modification, which I have ventured to call, in the
absence of a better word, diphthongisation, should



(delwedd B6107)


take place in "Welsh, Irish, English, or anv other
language. If you consult musicians on the matter,
they will tell you that a long and sustained note
has a tendency to lose its quality and change its
pitch : in other words, " there is naturally a great
difficulty in prolonging a sound at the same pitch
and with the same quality of tone," as Mr Ellis
ohserves in the fourth volume of his work on
Early English Pronunciation, p. 1273. He does
not dismiss the question without pointing clearly
to the source of the difficulty : "To retain the
vowel quality for a sensible time requires an un-
natural fixity of muscle, and consequently relaxa-
tions constantly occur, which alter the vowel
quality." Thus it turns out to be simply a ques-
tion of muscle, and the difficulty of prolonging
a vowel sound unmodified is exactly of the same
kind as that which one would soon feel in trying,
to iold one's hand up steadily for a length of
time, a method of torture which was well known
to Welsh schoolmasters when I was a boy.

The phonetic change here in question has justly
been called one of the great alterative forces
in language ; the latter, however, holds itself
free to have recourse also to the kind of change
exemplified in the reduction of diphthongs into
single vowels. Of this instances have already
been alluded to, as where Aryan ai and au were



(delwedd B6108)


supposed to have been reduced in the Goidelo-
Kymric period to S and o, also Aryan di into u,
whereby the contihuators of Aryan di and au as-
sumed the same form. But the common Goidelo-
Kymric antecedent of the- Welsh u to which 0.
Irish oi corresponds, may, as far as we now can
see, be presumed to have been ai or oi. As a
parallel to the reduction of Welsh oi into u may
be mentioned the case of Greek oi, which had in
the 11th century or earlier got to be sounded
like V — hence the habit of calling the latter v
y^iXoy, just as 6 was called e iln\oy when ai had
acquired its value — before its sound (y = ot) was
modified into that of t or t], as in the Greek of the
present day. I might dwell on the almost iden-
tical treatment of 0. Irish oi in Mod. Irish, where
the digraph ao has the sound of Welsh i, or
one between that and Welsh u. The English
and Latin parallels are less striking ; but if you
trace 0. Latin oinos to the more common forms
unus, una, unum, and down into the French un
une, the analogy between the history of the latter
and that of the Welsh un is in every respect very
close. The same kind of change is not unknown
in the dialects of Mod. Welsh : for instance, the
pronunciation prevalent in many, if not most,
parts of S. Wales of such words as doe, ' yester-
day,' oes, ' is,' traed, ' feet,' llaetA, ' milk,' is do, 6s,



(delwedd B6109)


trdd, lldth : so the e and y brought together by the
elision of a ^ form a modern diphthong liable to
be simplified as in tyrnas or ternas for teyrnas, ' a
kingdom,' and in Anglesey and Carnarvonshire
such plurals as tor/eydd, 'multitudes,' and jooz/eyt^rf,
' pastures,' become tor/ydd and por/ydd: so Lleyn,
the western third of the latter county, is ' now in-
variably called Llyn.

All the foregoing cases of reduction of diph-
thongs fall under the head of assimilation, which
has been noticed more than once on a former
occasion. Now there are other kinds of assimila-
tion which play a part in the vowel economy of
Welsh, but before they can be discussed to advan-
tage the nature of vowels must be studied more
closely than has hitherto been done here. Now
the vowels belong to the category of musical
sounds, and those who wish to study them as such
could not do better than begin by carefully reading
the first part of Professor Helmholtz's great work
on TAe Sensations of Tone as a Physiological Basis
for the Theory of Music, lately translated into
English by Mr. A. J. Ellis: also part H. 11.
of his Appendix xix. to Helmholtz's text, and
Chapter xi. of the fourth volume of his own work
already alluded to, On Early English Pronuncia-
tion, especially pp. 1272-1281. I find that the
best thing I can do is to copy here briefly their



(delwedd B6110)


views, as far as they serve to throw light oa Welsh

Sounds are distinguished into noises and mu-
sical tones, by which are not meant the intervals
of tones and semitones. The difference between
the former is that the sensation of a musical tone
is due to a rapid periodic motion of the sonorous
body, and the sensation of a noise to non-periodic
motions. The vowels, though they are of the for-
mer description, may, owing to the friction of the
breath against the parts of the mouth, contain an
admixture of noise, which it is the business of the
singer to eliminate. Musical tones in their turn
are distinguished by their force or loudness, by
their pitch or relative height, and by their quality.
Their force or loudness depends on the extent or
amplitude of the oscillations of the particles of the
vibrating body ; that is, the longer the distances de-
scribed by the said particles, as measured from their
position of rest, the louder the tones produced.

Their pitch or relative height depends solely on
the length of time each vibration occupies, or, as
it is more usually put, on the number of vibrations
made in a second: that is called the vibrational
number of the sonorous body, and the greater it is,
the higher the pitch of the tone it gives. Methods
have been invented for the reckoning of vibrations,
and it is found that, if they sink so low as about



(delwedd B6111)


30 per second, the ear can scarcely collect them
into a series : others follow one another with such
rapidity as to count by thousands in a second. In
other words, musical tones are roughly said to
rano-e between 40 and 4000 vibrations in a second,

o '

and to extend over seven octaves, while those
which are audible at all range between 20 and
38,000 a second, and extend over eleven octaves,
which will serve to show the marvellous capacity the
ear has of distinguishing sounds in respect of pitch.
Musical tones differ in quality, as when we
distinguish the human voice from the note of an
organ, although it may be of the same loudness
and pitch ; this is, further, said to depend on the
form of vibration, which, in its turn, may vary
indefinitely. For. example, it may be pendular or
resemble the swings of a pendulum, as in the case
of a tuning-fork ; or they may be like the motions
of a hammer which is uplifted by a water-wheel at
regular intervals, as in the case of a string excited
by a violin-bow. Mathematicians and physicists
classify musical tones into simple and compound,
without including in the latter term chords, which
they regard as composite tones. Leaving these
last altogether on one side, the only tones they
look at as simple are those produced by pendular
vibrations, and all others they analyse into pen-
dular ones. This resolution of all other vibrations



(delwedd B6112)


into pendular ones was in the first instance
arbitrary and a mere matter of convenience, but
Helmboltz and others have shown that it has a
meaning in nature, and they consider it as proved
that the organism of the ear is such that it per-
ceives pendular vibrations alone as simple tones,
and resolves other periodic motions of the air into
a series of pendular vibrations, hearing the simple
tones which correspond to these simple vibra-
tions. Thus when a tone is produced, say c, on
the violin, a practised ear hears not oAly c, but
also its -octave c\ the fifth of the latter g, the
second higher octave c" , and so on, as follows : —






0, C, g', c", e", g", b"l>, c'", d'", e'".

1, 2, 3, 4, 6, 6, 7, 8, 9, 10.

Here c, the lowest note, is the fundamental or
prime partial tone ; it is also generally the loudest,
and gives its pitch to the whole tone. C' is .the
first (harmonic) upper partial, and it makes twice
as many vibrations per second : g' is the second
upper partial, and makes thrice as many vibrations
as c: so with the others, which become fainter
and fainter the higher they go. It is to be ob-
served that any interference with the relative
force or loudness of any partial tone or tones is



(delwedd B6113)


Tecognised by the ear as a change of quality of
the compound tone ; and vice versa the quality of
a compound tone depends on nothing whatever
but the relative force of the partial tones : it is
important to keep this resolution, in the last resort,
of quality into considerations of quantity in mind
as we go on. The question of the composition of
tones has been also successfully attacked from
another direction ; for Helmholtz has been able to
produce given tones by means of suitable com-
binations of the simple tones of forks tuned to
the respective pitches of the partials they are to
stand for.

Another meaning which this resolution of musi-
cal tones has in nature appears in the phenomena
of sympathetic resonance. An instance or two
will explain what is meant by the term : — Gently
touch one of the keys of a pianoforte so as merely
to raise the damper, and then sing a note of the
corresponding pitch, forcibly directing the voice
against the strings of the instrument : the note
will be heard from the pianoforte when you have
ceased to sing. When the strings of two violins
are in exact unison, and one is excited by the bow,
the other will begin to vibrate. It is well known
that bell-shaped glasses can be put into violent
motion by singing their proper tone into them.
Lastly, the vibrations of a fork which, has been



(delwedd B6114)


struck are rendered more strongly audible by
being held near the mouth of a bottle or any
other resonance chamber in which the air is of the
same pitch as the fork. As to the pitch of the
air in a bottle, anybody, however dull he may be,
may experiment on that : for instance, if you blow
over the mouth of a bottle when it is empty, you
will find that it yields a deeper and more hollow
sound than when it has been half filled with
water, and that its pitch -will be still higher when
it is filled nearly up to the neck.

In the case of the voice, the tones are produced by
the vocal chords in the larynx, and they are of the
compound nature already described ; and the cavi-
ties lying between the yocal chords and the lips
form one or more resonance chambers by which
the tones produced in the vocal chords are in-
fluenced. The mouth in speaking assumes a great
variety of shapes, and as many of the latter as
imply also a difi"erence of pitch of the resonance
chambers they form will exercise a difierent in-
fluence on the quality of the tone ; for resonances
differing in pitch reinforce different partial tones,
which is at once recognised by the ear as a
change of quality of the compound tone. When,
for. instance, the resonance cavity of the mouth is
at its full length in ordinary pronunciation, its
pitch is lowest, and it reinforces the prime partial



(delwedd B6115)


tone, which then yields our w (English od) : com-
pare the case alluded to of the empty bottle.
When the same resonance cavity is at its shortest,
and its pitch, consequently, high, it reinforces the
very high partials, and the vowel produced is
Welsh i : compare the case of the bottle filled with
water nearly up to the neck. An intermediate state
of the resonance causes the reinforcement of some
of the lower partials, thus producing our a : com-
pare the case of the bottle half filled with water.
Of course the pitch of the tone is here assumed to
be constant as produced by the vocal chords, and
the pitch of the resonances to vary : it is to this
variation that we owe all the tone-qualities which we
write in Welsh a, e, i, o, u, w, and to nothing else.
Professor Helmholtz has succeeded in com-
pounding the tones of the more common vowels
from the simple tones of tuning-forks, thereby
also assigning the relative force of the different
partials required to make up each vowel : in
other words, he can make his forks, which he
regulates by means of electricity, sing out the
German vowels a, e, i, o, u, which I roughly ven-
ture to treat as equivalent to our a, e, i, o, w.
Many experiments have been made by different
men to ascertain the exact pitch or vibrational
number of the resonance cavities for the vowels.
One of them has arrived at the following results,



(delwedd B6116)


• when the vocal, chords are tuned to ^j, and c' is
assumed to make 256 vibrations in a second : —

Vowel w, 0, a, e, i.

Note b^, b\, b\ b"\, b"\

Vibrational No.... 224, 448, 896, 1792, 3584.

According to this, the pitch of the resonance
implied in the vowels rises an octave successively
in the order here given : unfortunately, this simple
relation is not corroborated by the experiments of
other investigators. However, they do not so far
differ as to establish another order of the vowels,
though they do not find the intervals to be ex-
actly the same. It will suffice for our purpose to
assume, what is fully sustained by the present state
of the evidence, namely, that the difference of re-
sonance pitch between m and a is greater than be-
tween 70 and or and a, and so with the others.
In other words, I would say that the vowels w, o,
a, e, i, are separated each from the next to it by
a single step, without insisting on the four steps
being exactly equal.

Should it, then, be found that w coming near
a is modified into o, or a coming near i is modified
into e, these and the like would clearly be cases
of partial assimilation. Now assimilation of this
description is well known to be a marked feature
of the Finnic languages, but it is not unknown in



(delwedd B6117)


other languages, and among them in Irish and
Welsh. The Irish instances have been discussed
at some length by Ebel in Kuhn's Beitraege in the
course of his Celtic studies in the first volume of that
publication, I will confine myself to a brief men-
tion of a few of the Welsh ones. Foremost among
the latter may be mentioned the sequence u — a,
making o — a in the history of simple adjectives
such as these : crmm ' curved, bent,' fem. crom^
erwn ' round,' fem. cron, dwfn ' deep,' fem.
dofn, Jmn ' this,' fem. hon, llwm ' bare,' fem.
Horn, and trwm ' heavy,' fem. trom. Now trwm,
trom, for example, points to a common Celtic pair
of forms, trumba-s mas., trumbA fem., which be-
came respectively in the course of time trumb and
trumba, the ending of the masculine having been
discarded earlier than that of the feminine, which
is supposed to have retained it until the a had
caused the u to be assimilated into o, whereby
trumba became tromba : lastly the a disappeared,
but not without thus leaving the feminine of the
adjective a form distinct from the masculine.
Trwm, I may notice in passing, is of the same
origin as the English verb to throng and the Ger-
man drang and druck, the b of the trumb- it im-
plies being the regular Celtic continuator of gv,
which is attested in the 0. Norse throngva, * to
press.' In the case of pwdr, 'rotten,' fem. podr,



(delwedd B6118)


the Latin adjective, from which these words are
borrowed, seems to have been treated as though it
were not putris, but putrus, putra. It is not to
be inferred from these instances that the assimila-
tion in question is confined to adjectives : most
Welsh names of the feminine gender which
happen to be monosyllables with the vowel o are
illustrations of it. In a few cases a form with
m has been suggested by that in o : thus from
Latin furca we have fforch and also ffwrch, but
both feminine : ffordd, ' a way,' yields the phrase
iffordd, 'away,' which is iffwrddin South Wales :
so also cwd seems to be later than cod, which,
though differing in gender, have the common
meaning of the word bag. This much by way of
introduction to a word of considerable interest :
Venantius Fortunatus, a travelled Italian of the
6th century mentions, among other musical in-
struments known in his day, a " chrotta Britanna."
This chrotta, which I take to be his spelling of
crotta, is in point of form the prototype of our
modern word croth, feminine, and in point of
meaning of the masculine crwth ; croth now means
the womb, also the calf of the leg, while crmth
means the crowd or rote, a box hollowed out of a
piece of wood especially for holding salt, and a
hump on the back. So, unless there were crutt
alid crotta synonymous in meaning, which is cer-



(delwedd B6119)


tainly very possible, one must conclude that crotta
had all the meanings mentioned, that is to say,
until it suggested a corresponding masculine to
share them with it. This view is confirmed by
the fact that the Irish form cruit remains feminine,
and means both a crowd or fiddle and a hump on
the back. The crwth was undoubtedly so called
from it shape, and the word for it appears to be of
the same origin as the Greek Kupro?, /cw/arr), Kvprov,
'curved, arched, round, humped, convex!'

Similarly among the instances of the sequence
i — a making e — a, the gender adjectives claim
the first place ; the following are some of them :
bryck ' fveckled,' fem. brech, hyr ' short,' fem. ber,
crych ' crisped,' fem. crech, gnlyb ' wet,' fem.
gwleb, gmych ' brave, fine, noble,' fem. gweck,
llym ' sharp,' fem. llem, melyn ' yellow,' fem.
melen. Here brych, brech, for instance, stand for
bricc, brecca = bricca ; but I hesitate to include
in the same category the adjective gnyn^ ' white,'
fem. gwen, the antecedents of which may have
been not vind, venda, but vend, venda, for the
Breton form is gwenn of both genders, and while
the syllable vend occurs several times in our early
inscriptions, vind is unknown in them. In this
case the assimilative action of the a of the feminine
would have been simply negative, with the effect
of preventing the e passing into y as in the



(delwedd B6120)


mascnline. To the foregoing may be added one or
two adjectives from Latin, such as ffyrf, ' strong,
stout, solid,' fern, fferf, from firmus, firma; and
sych, 'dry,' fem. sech, from siccus, sicca; nor are
there wanting instances of nouns such as cylcked,
' a bedding or bedcover,' from culcita, irmneg, ' a
glove,' from manica, and gramadeg, ' a grammar,'
from grammatica. There is, however, a native
Welsh ending eg = -ica, as in daeareg, ' geology,'
from daear, ' earth,' and Cymraeg, ' the Welsh
language,' for some such a form as Combragica, the
masculine being Cymreig, ' Welsh,' for Combragic.
There are also in use in Welsh the feminine termi-
nations ell (=-illa) and es {=-issa or -ista), as in
the case of priddell, ' mould, clod,' from pridd,
' soil, mould,' brenhines, ' a queen,' from brenhin,
' a king.' And one of the most useful terminations
in the language is en { = -inna or -inda), which is
matched in the masculine by -yn, as in melyn,
melen, ' yellow ' : take as examples cloren, ' a
tail,' from clarsr, ' covering, a lid,' dalen, ' a leaf,'
plural dail, seren, ' a star,' plural ser.

There now remains the converse change of a — i
into e — i, which takes place indifferently where
the i remains and where it is blunted into y, as
in the following instances: — Cyntefig 'pristine,'
from cyntaf ' first,' glendid ' cleanness,' from glan
' clean,' keli ' brine,' from hal-en ' salt,' iechyd



(delwedd B6121)

ńLECTURE in. 121

' health,' from iach ' healthy,' plentyn ' a child,'
from plant ' children,' rheffyn * a cord or rope/
from rhaff ' a rope ; ' these last belong to that
extensive class of formations already referred to
apropos of the ending en of the feminine.

Further, the passing of a into ei — liable in
Mod. Welsh to become ai — has commonly been
attributed to the effect of an i; but this is not
quite correct, for the occasion of the change is not
the presence of the yowel i, but of the semi-vowel
so written in "Welsh, which it will here be ex-
pedient to write j. The correctness of this view
will appear to any one who is content to proceed
from the known to the unknown. When the
Welsh borrowed Latin words, they seem to have
treated Latin i unaccented and followed by an-
other vowel as _;' ; so we have breich (now braich),
' the arm,' from brachium; rhaidd, ' a spear or pike,'
from radius, ' a staff, spoke, beam ; ' cyd-hreiniog,
* feeding together,' from prandium, ' breakfast, the
fodder of animals ; ' rheii^o, ' to snatch, bewitch,'
from rapio, ' I seize, carry off, ravish, captivate ; '
yspaid, ' a space of time,' from spatium. Simi-
larly, Maria and Daniel, treated as dissyllables,
yielded in Welsh Meir (now Mair) and Deinjoel
(now DeinjoV). So in native words such as
lleiddjad, ' a slayer,' from lladd, ' to kill,' edifeir-
jol, 'repentant,' iiom edifar, 'sorry for, full of



(delwedd B6122)


regret for,' creijjon, ' scrapings,' from crqfu, ' to
scrape,' and meibjon, ' sons,' from mob, ' a son.''
Thus it seems natural to conclude that such forms
as ffeir (now ffair), ' a word,' stands ior gar-j-, with
a termination — perhaps ja — which began with j,
but which has altogether disappeared excepting
that the j constantly reappears in related or de-
rived forms, such as, for instance, in the case of
gair, the plural geirjau, ' words,' or the derivative
geirjad, ' a wording.' This cajjegory would include
a very large number of words, and among others
such plurals as brein (now braiii), ' crows,' from
bran, ' a crow,' and the old neuter plurals of
which the 0. Welsh enuein, ' names,' may be
taken as a specimen — this and the 0. Irish plural
anmann seem to point to a lengthened form, an-
man-ja. Possibly, also, such third persons singu-
lar of the verb as geill, (^ he, she, it) can,' stands
for galljat (= galja-ti), -with, which compare the
Lithuanian galiu, ' I can.' The assimilation in
all the examples here enumerated must have at first
consisted in replacing the sequence a—j-, hye—j- ;
further preparation for the_; was made by making
the latter into ei—J-. In Breton and Cornish this
second step was never taken ; hence it is that
to our breicA and geir they oppose brecA and ger.
But this is not unknown in Welsh itself : thus in
the Liber Landavensis, BrycAeinjog ' Brecknock-



(delwedd B6123)


shire ' is called Brechenjauc, from Brychan's name,
and the name Meirckjon is there mostly given as
Merchjon or Merchjaun, supposed to be the Welsh
forms of the Latin Marcianus ; nay even now cen-
jog and celjog may be heard in Denbighshire,
Anglesey, and probably other parts of North
Wales, for ceinjog, ' a penny,' and ceiljog, ' a cock.'
In a few instances o — -j- also becomes e — -j- and
ei — -j~, as in yspeil {now yspail), 'spoil,' from Latin
spolium, and Emreis (less usual than Emrys),
from Ambrosius. I have not yet observed any
native instances in point. And where the original
sequence was e—j-, we sometimes find it super-
seded by ei — ;;;-, as in tdrthon, ' the tertian ague,'
from Latin tertiana, and in unbeinjaeth, which is
sometimes to be met with for the more usual
unbennaethy ' monarchy,' and in North Wales,
heddyw, ' to-day,' has passed through keddjm into
heiddjw, which is the prevalent pronunciation of
the word there at the present day.

As it is beyond the scope of this lecture to fol-
low the Welsh vowels into all their details, atten-
tion will now be directed to a number of changes
which amount to a reorganisation of the whole
system. But a few words must be premised on
the tone or syllabic accent in Welsh, and the
quantity or force of the vowels as regulated by it
and the consonants immediately following them.



(delwedd B6124)


Welsh monosyllables have an independent accent
with the 'exception of about a dozen proclitics.
The great majority of longer words are paroxy-
tones, and most of the exceptions are more ap-
parent than real, being perispomena, such as
glanhdu, ' to cleanse,' from glanhd-u = glan-
ha{g)-u, and cyfjawnhdd, ' justification,' from cyf-
jamnhohod = cyfjawnha{g)-ad. Moreover, a few
oxytones may still be heard, such as ymolch,
' wash thyself.' In 0. Welsh, words accented on
the final syllable seem to have been much more
numerous than now, and to have included all
words which had the diphthong aw (au) in it :
take, for instance, Aestaur, ' a sextarius, a measure
of capacity,' bardaul, ' bardic,' and the like.
Welsh vowels, when single, admit of being pro-
nounced in three ways — they may be either long
or short, and, when short, they may be either
open or closed. It will suffice to call them long,
short, and closed respectively. The long vowels
are much of the same quantity as in English :
thus our bod is pronounced like English bode with
long 0. The short vowels also occur in both lan-
guages : the i, for instance, of dinas, ' a city,' and
and the y of myned, ' to go,' sound very nearly
like the English i and o of dinner and money
respectively. The closed vowels are those which
are suddenly and forcibly broken off or closed by



(delwedd B6125)


a succeeding consonant : our pen, ' head,' tan,
' under,' at, ' to,' sound in this respect like the
English words pen, tan, at. A word now as to
their distribution : accented monosyllables have
their single vowels long or closed, short ones
being admissible, only in the proclitics. Longer
words, which are not perispomena, admit only
short and closed vowels : short or closed in the
tone-syllable, short only in other syllables ; and,
conversely, all unaccented syllables have their
single vowels short. These distinctions have
regard only to the quantity and force of. the
vowels, not to. their quality ; for although k good
ear could hardly fail to detect differences of qua-
lity between the a's, for instance, in tan, ' a fire,'
tdfiau, ' fires,' tdnjo, ' to. fire,' the language treats
them as the same a varying in quantity and force,
and so they will here be dealt with.

The triple pronunciation of the vowel is, as it
has just been pointed out, recognised in English,
but in Welsh it has been stereotyped into a sys-
tem, the meaning of which it is the business of
phonology to explain. The vowels of the Aryan
parent-speech may be regarded as having come
down into Early Welsh with values which may,
roughly speaking, be called constant, whereas the
value of those of Mod. Welsh, as far as regards
their quantity and force, depends on their posi-
tion. The question, then, is how they came to



(delwedd B6126)


exchange their constant values for positional
values, and how comparative uniformity was
elicited from the original variety. The cases to
be taken into account range themselves into
three groups : those where long vowels have been
shortened, those where short vowels have been
lengthened, and those where no perceptible change
of force or quantity is attested.

Take the first : that a long vowel should be short-
ened when it occurs in an unaccented syllable seems
to us, with our modern way of marking the accented
syllable by a greater stress of the voice, so natural
as to require no remark, and we pass on to the
same modification when it happens under the
accent. This concerns the vowels u, i, and the
Early Welsh continuator.of Aryan a. Thus u is
shortened in unol, ' united,' and closed- in undeb,
' union,' from un, ' one,' and so in other words.
Traces of the operation of this law, which is
general in "Welsh, may be found in English ;
witness such words as nose, nostril; vine, vine-
yard ; house, husband, hussy ; nation, national.
It is not, however, confined to these more palpable
cases, for Mr. Barlow finds that the syllable ex,
for instance, when pronounced by itself, appears
in the diagram described by the marker of the
logograph considerably longer than when it is
spoken as a part of such a word as excommuni-
cate; in the latter it becomes, he says in the



(delwedd B6127)


paper already alluded to, compressed, its length
being shortened and its height increased. The
reason for such a law is perhaps to he sought in
the fact that the centre of gravity, so to speak,
•of a word is in the accented vowel : if that hap-
pens to he in the final syllable, it may remain
long ; if not, there seems to exist a sort of in-
stinctive tendency to share the breath and time
required for uttering that syllable between it
and the remaining portion of the word. The
ideal limit of this would be to devote exactly
the same amount of breath and time to the pro-
nunciation, for instance, of tanau and tan, of
national and nation. The comparatively rare oc-
currence of such cases of vowel-shortening, due
to the influence of the accent in Latin, still rarer
in Greek, as well as the nature of the metres the
Greeks and Eomans used in their poetry, seems
to warrant the inference that the ancient accent
mainly implied a difference of pitch, while ours
in English and Welsh mainly means a difference
of loudness or force, the change of pitch being
mostly considered secondary, or passed over un-
observed. As we go on it will appear by no
means improbable that "Welsh was adopting (or
had already adopted) in the 8th century our
modern accent in lieu of that which may be called
the classical accent. The effects of such a change



(delwedd B6128)


must have been very considerable on our vowel
system, though they are exceedingly hard to de-
fine. But as similar changes have occurred in the
history of the majority of the modern languages
of Europe, comparative phonology may reasonably
be expected at some future day to solve the pro-
blem satisfactorily.

The next vowel is i, which we failed to detect
as the continuator of Aryan i. It is even doubtful
whether it was not sometimes z in Early Welsh, as
well as i. It would be hard, for instance, to prove
that it was at any time long in the word elin:
the cognate forms are Ir. tiille, " ulnas," Eng. ell,
el-bow, Lat. ulna, Greek aiKevr}, Skr. aratni; and
it is certain that it never was long in anifel, ' an
animal,' from Lat. animal or one of its oblique
cases. However, even where it must have always
been long in Welsh, as in gmr, ' true ' (Ir. fvr,
Lat. virus), and dm, ' a fort, a town ' (Ir. dun,
Eng. town), we find the quantity of the vowel
short when a syllable is added, as in anwiredd,
' untruth,' and dinas, ' a city,' and so in others.

.The fortunes of Aryan a in Welsh are still
more interesting : towards the close of the Early
Welsh period it had become o, which by the 9th
century had been diphthongised into aw (written
au) in monosyllables and other words where it
was accented in the final syllable, as in 0. Welsh



(delwedd B6129)


lau, now llaw, ' a hand,' and paup^ now pawb,
' everybody/ and the like ; but in those positions,
where long vowels are inadmissible, not oijily
was its diphthongisation into aw arrested, but the
was reduced sooner or later to o: so by the
side of paup and hestaur (sextS,rius) 0. Welsh
offers us popptu, ' on every side,' and hestorjou,
the plural of kestaur, and so on. So it seems
probable that the reorganisation of the Welsh
vowel system came upon the vowel in question
when it was 5, but before it had begun to be
diphthongised into aw. In Bede's liistoria Eccle-
siasiica, as edited by Mr. Moberly (Oxford, 1869),
the proper names have been printed as they occur
in the oldest manuscript of the work, which is
assigned to the year 737, and there the Abbot of
Bangor who met Augustine is called Dinoot.
Welsh tradition calls him Dunaut, later Dunawd.
There can be no doubt as to the virtual identity
of Dinoot and Dunaut, nor, as I think, as to both
being forms of the Latin name Donatus, which
was not unknown in Britain in the time of the
Eoman occupation, when many more Latin names
were adopted by the Britons. Now Dinoot and
Dunaut show that Bede had the same diflSculty in
distinguishing Welsh u from I as the natives of
South Wales have in our own day, and that his
00 probably meant o, which had not been diphthono--




(delwedd B6130)


ised. Bede is supposed to have lived from 672
to 734, but he may have been copying from an
earlier writer. However, we should probably not
be far wrong in supposing the- reorganisation of
the vowel system to have been in process during
the century from 650 to 750 : probably it began
long before, and it is certain that it lasted long
after. It is worth while observing, that the same
law which gives us au in monosyllables and o in
longer words, has also been at work in Irish, as
in the following words, which I copy from the
Gram. Celtica^ p. 18 : — cliah, " corbis," cUbene,
" sporta ; " Jiach, " Aebitam" fechem, "debitor;"
grian, "|Sol," grene, "solis;" sliah, "mons," slehihj
" montibus," to which I would add dia, ' god,'
genitive dii for divi. In the case of ua and 6
more uncertainty prevails, but Zeuss (p. 23) gives
huar, '•' hora," genitive hore, and suas, " sursum,"
but i sosib " in altis."

Next comes the group which comprises the
cases of vowels undergoing a lengthening. This
happens almost exclusively in monosyllables, and
conversely it takes place in all monosyllables —
provided they are not proclitics, or that their
vowels are not already m, I, or a diphthong —
which close with any one of the consonants g, d, b ;
dd,f; and n and I, where they were not formerly
doubled or accompanied by another consonant.



(delwedd B6131)


Take, for instance, the following -words : gmag,
' empty,' tad, " father,' pib, ' a pipe,' hedd, ' a
tomb,' claf, ' ill,' glan, ' clean,' pwl, ' blunted ; '
if the word is lengthened by the addition of a
syllable, then the vowel returns to its original
quantity, as in beddau, ' tombs,' and glanach,
' cleaner.' This process of lengthening the vowels
of monosyllables was not complete in the early
part of the 0. Welsh period : witness the Capella
glosses hepp, now Mb or eb, ' quoth,' and nepp,
now neb, ' anybody.' Neither is it easy to ac-
count for ; but it may be surmised that, as most
of our monosyllables represent words originally of
two (or sometimes more) syllables, the vowel of
the leading syllable was reinforced by way of
compensating for the discarding of the rest of the
word, a long monosyllable being, metrically speak-
ing, a better equivalent for a dissyllable than a
short one. Possibly, also, the mistaken analogy
of such forms as paup and popptu exercised an
influence in the same direction. There is another
consideration which is of more weight than the
foregoing : in the earlier stages of the Aryan lan-
guages the pitch-accent prevailed, and conse-
quently a mode of pronunciation was usual which
is far less so in those of their modern repre-
sentatives, where the stress-accent is dominant.
I allude to such words as Latin pater, bonus.



(delwedd B6132)


These were not patter, bonnus, in spite of the
French bon, bonne, nor pater, bonus, in spite of
the Italian padre, buono, and the Anglo-Latin
monstrosities payter, bownus. But enough has
been said to show that such a word as bonus had
a tendency, under the influence of the stress-
accent, to become either bonnus or bonus. The
latter represents the course with which the student
of Welsh is mostly concerned. The same ten-
dency is well known also in Modern Greek, where
Xoyo? is now Xayo?, and it is widely stereotyped
in Mod. High German, which is said to be dis-
tinguished from. Mid. H. German by its lengthen-
ing the short tone-vowels followed by single con-
sonants, as in geben, ' to give,' and haben, ' to
have.' We have it also in English : take the
words ape, make, late, lame, which were formerly
apa, macian, lata, lama. The analogy between the
English words and the Welsh ones in question is
so complete — both lengthen the tone-vowels, and
both discard the inflectional endings — that one
cannot help suspecting their having been subjected
to the operation of the same causes.

In the foregoing enumeration of the consonants
requiring long tone-vowels to precede them, no
mention was made — the explanation required being
somewhat different — of the rule, that the vowel
must also be long before ch, th,ff, and s, as in



(delwedd B6133)


cock, ' red,' crotA, ' the womb,' rhaf, ' a rope,' and
fflas, ' blue, green, grey.' The antecedents of
these spirants were respectively cc (or cs), tt, pp,
and ss (mostly for st) : take for instance our cock,
which is probably from coeeum, ' scarlet,' and crotA,
which has already been traced to' crotta : these
were no doubt pronounced coccum and crotta, which
might be expected to have yielded in the first
place cock and crotL These last would eventually
become each and crotk, owing to the analogy of the
other cases already mentioned, and to the reaction
on the vowels of the spirants, which, not being
instantaneous in their pronunciation, are not
favourable to a clean cutting off of the vowels
preceding them. And so in the case of the other
spirants, including s, whence a difference between
Irish and Welsh in words otherwise identical,
such as fflas ; ours being fflds, while the Irish is
fflas. Supposing the steps coccum, cock, coch were
made out, we should still find a difficulty in as-
signing the time when the .short vowel was
lengthened ; but Welsh verge offers a case of
assonance which deserves a passing mention.
Dafydd ab Gwilym (1340-1400) makes och, ' oh,'
answer such words as cocA, /ed,' and clocA, ' a bell,'


" Och ! Ooh ! y Ddol Goch wedi gwyl."

Now the interjection is an exception, being pro-



(delwedd B6134)


nounced not och but och, and such assonances have
been supposed to show that its pronunciation was
formerly regular, that is och. But the question
may be put in two ways : has och been shortened
contrary to analogy, or has it merely retained its
original quantity of vowel contrary to analogy?
In the latter case it would follow that D. ab
Gwilym spoke clock, cock, and not clock, cock, as
we do.

So far of the vowels which change their quan-
tity, and of the conditions under which that
happens : a word now on the third group, where no
perceptible change of quantity has taken place.
The instances here in point are of two kinds : words
with closed vowels as bdlck, ' proud,' bdlchder,
' pride,' plant, pldntack, ' children,' darn, 'apiece,'
ddrnau, ' pieces ; ' and those with short vowels
such as kanes, ' history,' qfal, ' an apple,' maddeu,
' to forgive.' In these no great change of quan-
tity of the tone-vowels can have occurred from
the earliest times, though no doubt some modifica-
tion may have followed the passage from the
pitch-accent of the ancients to the stress-accent
of our own day. The number of instances in this
third group is probably far in excess of that in
the two former groups put together, if we confine
ourselves to the tone-syllable, which after all is the
kernel of all our words : so that our vowel system



(delwedd B6135)


has altogether been more conservative than might
be inferred from the somewhat lengthy remarks to
which those other groups gave rise.

The processes already mentioned of reorganising
the Welsh vowel system were probably well over
by the end of the Mediteval Period in the history
of the language. Before concluding- this lecture
a few more have to be noticed, some of which are
not only later in time than the foregoing, but, to
some extent, probably owe their origin to the
influence of the analogy of the latter. Consider
for a moment the individuality so strongly im-
pressed in the ways already pointed out by Welsh
phonology on certain monosyllables as compared
with the same when forming parts of longer words,
and take as instances the following : — coch^ ' red,'
superl. cochaf, llath, ' a rod,' llathen, ' a yard,' tad,
' a father,' tddol, ' fatherly,' mdb, ' a son, a boy,'
mebyd, ' boyhood,^ brawd, ' a brother,' brodyr,
' brothers,' tawdd, ' molten,' toddi, ' to melt.' Here
we have a tolerably well-defined contrast which
came to be impressed on another class of words,
namely, such as have a diphthong in the tone-
syllable. This was done by adding, so to say, to
the weight of the monosyllable, by diminishing
that of the corresponding part of the longer form,
or by both processes at once. The diphthongs, the
history of which is here concerned, are our modern



(delwedd B6136)


ai, au, ae, oe, Tcy. Mediaeval Welsh ei becomes ai
in modern monosyllables, as in bd, now hai,
' blame,' pi. beiau, geir, now gair, * a word,' pi,
geirjau, Meir, now Mair, ' Mary ; ' tbe proclitics
ei, ' his,' ei, ' her,* are of course not subject to
this change : the same applies to independent
monosyllables which happen to be already sufr
ficiently weighted, as when they end with two
consonants, such as gei/r, ' g6ats,' meirch, ' steeds,'
ysceifn, the plural of yscafn, ' light, not heavy.'
Med. Welsh eu becomes au, as in deu, now dau,
' two,' and keul, now haul, ' sun,' heulog, * sunny ; '
the proclitic eu ' their ' remains, like ei, un-
changed : the same applies to neu, ' or.' Old
Welsh ai (pronounced probably with the blunted
i, which we now write y or m) becomes ae so early
as the beginning of the Med. Welsh period, as for
instance in air, later aer, ' a battle,' and cai, later
cae, 'a field.' The spelling ae, however, is also
retained in words of more than one syllable, as in
aerfa, ' a battle-field,' and caeau, ' fields. But
the pronunciation varies between au or ai and eu
or ei. In a few words this relation is optionally
indicated by the ordinary orthography, as in aetk,
' ivit,' but euthum, ' ivi,' and euthost, ' ivisti,'
maes, ' a field,' meusydd, ' fields ; ' in the collo-
quial, ae in an unaccented final syllable is mostly
reduced into a single vowel, whereby such words



(delwedd B6137)


as hiraeth, ' longing,' become in South Wales
hiretk, and the like. A word which in 0, Welsh
would have had the single form mat, is in Mod.
Welsh both mae and mat : the former means ' is,'
the latter is a proclitic with the force of the Eng-
lish conjunction that : the same use of a verb as a
conjunction occurs in taw, ' that,' commonly used
in South Wales instead of mai : taw is obsolete as
a verb, but not so its Irish equivalent td, ' is.' 0.
Welsh 01 (also probably pronounced with i = our
modern u or y) makes oe in Med. Welsh, and
later, as when 0. Welsh ois becomes oes, ' age,
generation,' and oid becomes oedd, ' was.' The
spelling oe is also retained in other words than those
of one syllable : take for instance the 0. Welsh
ois oisoud, ' sseculum sseculorum,' later oes oesoedd,
pronounced in North and South Wales respectively
oes ousoudd, oes oisoidd, or still more colloquially
with ousodd, oisodd, the diphthong in the unaccented
ending being reduced to a single vowel as in many
other words, such as mynyddodd, ' mountains,'
nefodd, ' heavens, heaven,' written mynyddoedd,
ne/oedd. As to the diphthong 7vy, when it occurs
in an accented syllable followed by another syllable
in the same word, the accent under favourable
circumstances shifts from the w to the y, whereby
the former becomes a semi-vowel, as in gwydd,
'a, goose,' but gnyddau, 'geese.' This modi-



(delwedd B6138)


fication is probably very modern, and otherwise
this diphthong may be regarded as the most un-
changeable, excepting ew, in the language, as
the old spelling ui probably meant exactly the
same sounds which we write wj/. But as m/
and oe represent an early oi which came down
into 0. Welsh partly as oi (now oe), partly as ui
(our my), the difficult question as to the cause of
this bifiurcation meets us. The following answer is a
mere guess, to be taken for what it is worth. In
Mod. Welsh the diphthongs, when accented, have
the accent on the leading vowel (excepting in such
cases as that of gwyddau, where 7oy ceases to be
a diphthong), as in gdir, mde, oedd, and gwydd.
But it may well be that it was not always so, and
that gair, for instance, was preceded by geir for
geirja and garjd, the advance of the accent having
been gradual — garja, geirja, geir, gdir. Take also
such words as draen, ' a thorn,' plural drain,
which may be inferred to stand for drain sing.
drein plural, and these for dragn and dregn-i or
dregn-ja : the cognate Irish is draighen, ' thorn.'
Similarly dau would imply deu, and so in other
instances. Should these guesses turn out well
founded, one would have to regard oen, ' a lamb,'
for instance, and its plural wyn, as representing
oin sing, and oin plural, for oin-i or oin-ja, with
an ending indicative of the plural number retained



(delwedd B6139)


intact at a time when the singular had been re-
duced to a monosyllahle. This agrees tolerably
well with the fact that Latin e makes ny in Welsh,
as in canwyll, ' a candle,' and afwyn, ' a rein,' from
candila and habena, while the oxytone Aavir\K has
in "Welsh yielded Deinjoel, now Deinjol. If the
antecedents of our ai, au, ae, oe, ny were ei, eu,
di, 6i, 01, the modification thereby implied admits
of being described simply as the replacing the
unaccented vowel by a nearly related vowel of a
lower pitch of resonance, a principle the working
of which is, I am inclined to think, also to be de-
tected elsewhere in the language : for instance,
where Mod. "Welsh replaces eu in unaccented final
syllables by au, as in pethau, ' things,' fforau ' best,'
borau, ' morning.' Compare also the disuse of
enwiredd, ' untruth,' engyljon, * angels,' llewenydd,
'joy,' in favour of the forms anwiredd, angyljon,
llawenydd, and the like.

( 140 )



(delwedd B6140)


"As his craze ia astronomical, he will most likely make few con-
verts, and will be forgotten after at most a passing laugh from scien-
tific men. But if his craze had been historical or philological, he
might have put forth notions quite as absurd as the notion that the
earth is flat, and many people would not have been in the least able
to see that they were absurd. If any scholar had tried to confute
him, we should have heard of ' controversies ' and ' differences of
opinion.' " — The Satuedat Ebview. .

It is my intention now to call your attention to
the continuity of the Welsh language ; but before
we attempt to trace it back step by step to the
time of the Eoman occupation, it may be well to
premise that history fails to give us any indica-
tions which would lead us to infer that the Welsh
of the present day are not in the main the lineal
descendants of the people whom the Eomans found
here. No doubt the race received an infusion of
foreign blood in those neighbourhoods where the
Roman legions had permanent stations ; but its
character ddes not seem to have been much in-
fluenced by contact with the English, at any rate
previously to the Norman Conquest. As to the
Danes, they have hardly left behind them a trace
of their visits to our shores, and that the Irish
occupied any part of Wales for a length of time



(delwedd B6141)


still remains to be proved. Certainly the effects of
such an occupation, even were it established, on
our language -will be hard to discover. The monu-
ments to be met with in Wales and elsewhere in
the West of Britain alleged to belong to the Irish
will presently come under notice. Thus it would
seem that we are entitled to expect to find our
Welsh to have been continued without any violent
interruption from the common language of the
Kymric race in the time of Agricola, to which be-
longed not only Wales, including Monmouthshire,
but also Devon and Cornwall, a considerable por-
tion of the west and middle of England, nearly
all the north of it, and a part of Scotland. To
what extent the country was occupied by non-
Kymric races is a question which will occupy us as
we go on. Subsequently to the decisive battle of
Chester in 607, when the English succeeded in sever-
ing the Welsh of Gwynedd from their countrymen
in Lancashire and the North, the Kymric popula-
tion of the west of the island found themselves
cut up into three sections, the Strathclyde Britons,
those of Wales, and those south of the Bristol
Channel. As to the northern section, it was not
long ere English drove the old language off the
ground. In Cornwall it survived to differentiate
itself considerably from Welsh, and to become
extinct as a spoken language only in the last cen-



(delwedd B6142)


tury. In the middle section, that is, in Wales,
you need not be told that it is still living and
vigorous, though its domain is getting more and
more circumscribed. One may accordingly assume,
at any rate provisionally, that the Kymric people
of the North, of Wales, and of Devonshire and
Cornwall, spoke the same language till the end of
the 7th century or thereabouts ; so in writing on
early Welsh we claim the use of ancient Kymric
monuments, whether they occur in Wales itself,
in Devonshire, or in the vicinity of Edinburgh.
Of course one is not to suppose that within that
range there were no dialectic variations ; but they
were probably not such as to make themselves dis-
turbing elements within the compass of our early
inscriptions. The case is different when the latter
are compared with those of Ireland, the -linguistic
differences between the Kymric and the Goidelic
nations being of a far older standing ; but more of
this anon.

Hitherto it has been usual to divide the Welsh
language, historically considered, into three periods,
namely, those of Old, Middle, and Modern Welsh.
This classification was adopted at a time when
very little was known to glottologists respecting
our early inscribed stones, which mark out for
us two periods of the language to which, in de-
fault of a better, the term Early Welsh may be



(delwedd B6143)


applied. This, however, cannot be' done without
rendering Middle Welsh inadmissible; but, in
order to disturb the old terminology as little as
possible, the adjective Medimval may be used in-
stead of Middle. Having premised this much, we
proceed to parcel out the entire past of the langu-
age in the following manner : —

1. Prehistoric Welsh, ranging from the time when

the ancestors of the Welsh and the Irish could
no longer be said to form one nation, to the
subjugation of the Britons by Julius Agricola,
or, let us say, to the end of the first cen-

2. Early Welsh of the time of the Eoman occupa-

tion, from then to the departure of the Romans
in the beginning of the fifth century. ,

3. Early Welsh of what is called the Brit- Welsh

period, from that date till about the end of
the seventh century, or the beginning of the

4. Old Welsh, from that time to the coming of the

Normans into Wales in the latter part of the
eleventh century.

5. Mediaeval Welsh, from that time to the Refor-


6. Modern Welsh, from that epoch to the present




(delwedd B6144)


This would be the order to follow if one had to
produce specimens of the successive periods of the
language, but for our present purpose it will be
preferable to trace it back step by step from that
stage in which we know it best to the other stages
in which it is not so well known ; in a word, to
treat it as a question of identity. The lead, then,
is to be taken by Modern Welsh, which I would
distinguish into Biblical and Journalistic Welsh.
By the latter is meant the vernacular, which we
talk, and meet with, more or less touched up, in
most of our newspapers. It is characterised by a
growing tendency to copy English idioms, the
result no doubt of frequent contact with English,
and of continually translating from English. It
is right to add that the number of the books and
journals published in it is steadily increasing.
Biblical Welsh, as the term indicates, is the lan-
guage of the Welsh translations of the Bible, and
a number of other books, mostly theological, of the
time of the Eeformation and later, and it is still
the language in which our best authors endeavour
to write. This overlapping of Biblical and Jour-
nalistic Welsh in our own day will serve to show
that, when glottologists divide, for convenience'
sake, the life of a language into periods, one is not
to ask the day of the month when one period
ends and the succeeding one begins. Passing be-



(delwedd B6145)


yond tlie time of the Reformation, we come to the
Mediaeval Welsh of the Bruts or chronicles, so
called from the fashion, once common, of manufac-
turing a Brutus or Brytus to colonise this island,
and to give it the name of Britain : he was held
to have been a descendant of ^neas, and thus
were the Welsh connected with Troy. To about
the same time are to be assigned the romances
called the Mabinogion, which consist mostly of
tales respecting Arthur and the Knights of the
Round Table. Here also may be mentioned, as
belonging to the earlier part of the period, the
Venedotian versions of the Laws of Wales, which
Aneurin Owen found to be in manuscripts of the
12th century, and it is to the 12th that Mr. Skene
assigns the Black Book of Carmarthen in the Hen-
gwrt Collection, the property of W. W. E. Wynne
of Peniarth, Esq. : it contains the oldest version
extant of much of the poetry commonly assigned
to the 6th century. As to the language of this
poetry, it is certainly not much older, if at all,
than the manuscript containing it I have said
the language, for the matter may be centuries
older, if we may suppose each writer or rehearser
to have adapted the form of the words, as far as
concerns the reduction of the mutable consonants,
to the habits of his own time, which one might
well have done unintentionally, and so, perhaps,




(delwedd B6146)


without the matter being much tampered with.
For the details of this question I would refer you
to the fourteen introductory chapters in Skene's
Four Ancient Books of Wales : suffice it here to
say, that the poems ascribed to the Oynfeirdd or
early bards belong, as far as concerns us now, to
the Mediaeval period of Welsh, though the metre,
the allusions, and the archaisms, which some of
them contain, tend -to show that they date, in
some form or other, from the 9th century, if not

So far we have at our service abundance of
literature for all philological purposes ; but when
we pass the threshold of the 12th century, the case
. is no longer so, our only materials for the study
of Old Welsh being inscriptions and glosses,
together with a few other scraps in Latin manu-
scripts. The inscriptions here alluded to are the
later ones, written in characters which archfeolo-
gists call Hiberno-Saxon. As to the manuscript
portion of the materials, when a Welshman read-
ing a Latin author met a word he did not under-
stand, he ascertained its meaning, and wrote its
Welsh equivalent above it, between the lines, or
in the margin : so our Welsh glosses were pro-
duced. We have, besides, fragments of charters
and scraps of poetry filling up spaces which hap-
pened to be blank in the original manuscripts.



(delwedd B6147)


Most of them are ia Oxford and Cambridgie, and
one in Lichfield. Their dates are ascertained for
us by experts, and it is to the 9th century that
they now assign the oldest collection. Altogether
they are far under a thousand vords and contain
few complete sentences : so, while they leave us
considerably in the dark as to the syntax of the
language, they enable us to ascertain what phono-
logical and formal changes it has passed through
since the 9th century. Among other things, we
are placed in a position to watch the appearance
and gradual spread in it of the more interesting
consonantal mutations.

The next move backwards lands us in the Brit-
Welsh period of the language, for the study of
which we have, besides a few names in Gildas and
other writers of the time, a pretty good number of
epitaphs, but mostly written in Latin. This is
unfortunate, as the Kymric names they contain
have, in a great number of instances, their termi-

nations' Latinised. A few, however, are bilingual,
consisting of a Latin version in more or less
debased Roman capitals, interspersed occasionally

towards the close of the period with minuscules,
and of an Early Welsh version in Ogam. Several
of them will be noticed as we go on ; and I now
submit to you a list [this will be found in an
Appendix at the end of thevolume] of them, con-



(delwedd B6148)


taiuing all those which have not been reduced to
mere fragments of no special interest, or rendered
illegible by centuries of exposure.

As we pass back from the Brit- Welsh period to
the time of the Epman occupation, our data become
still more meagre. They consist (1) of a few
proper names which have been identified in
Ptolemy's Geography, the Itinerary of Antoninus,
Tacitus' Agricola, and other writings of that time,
and (2) . inscriptions scattered up and down the
country occupied by our ancestors. The number
of Celtic names. in these last is very considerable,
but we cannot be sure that they are in all in-
stances Kymric ; however, we may assume some of
them to be so if they are found at Caerleon (that
is, the Isca Silurum of the ancients), at York,
and other places in the North. They are mostly
epitaphs written in Latin, and beginning with the
usual Koman dedication to the Di Manes, but some
are votive tablets to local gods. Any one who has
an eye for Celtic names can pick them out at his
leisur-e in the Corpus Inscriptionum Latinarum,
published not long ago in Berlin, under the
superintendence of Professor Hiibner: the seventh
volume is devoted to those of Great Britain.

And now that we have thus rapidly scanned
the past of our language so far back as any the
slightest assistance is rendered us by ancient



(delwedd B6149)


authors and contemporary monuments, you may
ask, What about the question of identity pro-
pounded at the beginning of the lecture ? As
far as concerns Modern and Mediaeval Welsh, or
Medisevaland Old Welsh, there can be no question
at all, and we need not hesitate to assume the
identity of the Welsh language of the 9th century
with that of the 19th ; that is to say, the former
has grown to be the latter. Nor is there any
occasion at present to prove its identity in the 1st
and 6th century, though, it must be admitted, that
would, owing to the scantiness of our data, be only
less difficult than to establish the negative. At
any rate, we may wait until the latter has found
an advocate ; for it is not just at this point that
the chain of continuity has been suspected : the
links that are now and then challenged occur
between the 6th and' 9th centuries, and it is to
them that our attention must now be directed.

Here precedence may be granted to the difficulty
of those writers who fail to see how a language
once possessed of a system of cases could get to
lose them and appear in the state in which we find
the Old Welsh of the 9th century, which hardly
differed in this respect from the Welsh of our day.
These may be dismissed with the question. What
has become of the cases of Latin in the languages
of the Romance nations of modern times, such as



(delwedd B6150)


Italian, French, and Spanish, or how many of the
five or six cases formerly in use in English are
current in Modern English ?

Then there are those who will have it, that
Welsh can never have had cases, because it is, as
they imagine, nearly related to, or immediately
derived from Hebrew, which also has no cases.
Neither do literary ostriches of this class deserve to
be reasoned with, at any rate until they have taken
their heads out of the sand and acquainted them-
selves with the history of the philological world since
the publication of Bopp's Comparative Grammar.
As matters stand, it would in all probability be use-
less to tell them that Welsh has nothing to do with
Hebrew or any other Semitic tongue. It is, how-
ever, not a little satisfactory to read, from time to
time, in the English papers, that this Hebrew
nightmare, which has heavily lain, some time .or
other, on almost every language in Europe, seems
to be fast transforming itself into a kind of spirit
of search impelling gentlemen of a certain idiosyn-
crasy to turn their thoughts to the .discovery of
the Lost Ten Tribes of Israel.

Not to dwell on the fact that Semitic scholars
are satisfied that Hebrew itself once had cases, or,
rather, that it never lost them altogether, it may
be interesting to notice that even the Welsh we
speak may be made to yield us evidence of the use



(delwedd B6151)

ńLECTUKE rV. 151

of a system of cases in the language during the
earlier periods of its history. But before we pro-
ceed to this we may for a moment consider what
traces of the cases of Latin remain in the Welsh,
words which our ancestors borrowed from that
language. Well, if you look through a list of
these loan-words, which amount in all to no less than
500 Latin vocables, you will find that some show
traces of the Latin nominative, as for instance,
lendith, ' a blessing,' ffnrn, ' an oven,' pabell, ' a
tent,' from benedictio,fornax,papilio, respectively,
while others are supposed to be derived from
accusatives, such as cardod, ' alms, charity,' ciw-
dod, ' a tribe,' j)ont, ' a bridge,' from caritatem,
cimtatem, and pontem: compare lorddonen, 'Jor-
dan,' and Moesen, ' Moses,' from 'lopBdvrjv and
MouvffTJv. Lastly, it may be left undecided whether
tymp, ' a woman's time to be confined,' comes
from tempus nominative or tempus accusative, and
so of corf, ' a body,' from corpus, but tymmhor,
' a season,' must have come from temporis, tempori,
or tempore, and so of the corffor in corj^ori, ' to in-
corporate,' and in corjvroedd, an obsolete plural
of corj^, for which we now use cyrf. Now, have
we any such traces in Welsh words of Welsh
origin ? No doubt we have ; and they are to be
detected by comparison with other languages,
especially Irish. The following are found to be
nominatives : —



(delwedd B6152)


bru, ' womb : ' compare 0. Ir. nom. bru, gen,

car, ' a friend : ' compare 0. Ir, nom. cara, gen.

ci, ' a dog : ' compare 0. Ir. nom. cu, gen. con.
gof, ' a smith : ' compare 0. Ir. nom. goba, gen,

llyg, ' a field-mouse : ' compare 0. Ir. nom. luch,

gen. lochad.
tan, ' fire : ' compare 0. Ir. nom. tene, gen. tened.

In other instances the comparison shows us
that the Welsh forms are not nominatives, but
probably accusatives, as in the following, pointed
out to me by Mr. Stokes : —

bon (in henfon), ' a cow : ' compare 0. Ir. accus.

boin, nom. bo.
breuan, ' a handmill : ' compare 0. Ir. accus.

broinn-n, nom. broo, equated by Mr. Stokes

with the Sanskrit grdvan, the Rigveda word

for the stone used in sq;ueezing out the- soma

breuant, ' the windpipe : ' compare 0. Ir. accus.

brdigait-n, nom. brdge.
dernydd, ' a druid : ' compare 0. Ir. accus.

druid-n, nom. drui {drym would seem. to be

the Welsh nominative).



(delwedd B6153)


emi7i, ' a nail of the hand or foot : ' compare 0. Ir.

accus. ingin-n, nom. inge.
gorsin, ' a door-post : ' compare 0. Ir. accus.

ursain-n, nom. ursa.
Iwerddon, ' Ireland : ' compare 0. Ir. accus.

Herenn, nom. Hiriu,
mis, ' month : ' compare 0. Ir. accus. mis-n, nom.

pridd, ' earth, soil : ' compare 0. Ir. accus. creid-n,

nom.- cri.

Add to these the word nos, ' night,' a nominative
for nots = noct-s : compare Latin nox, gen. noctis.
If Welsh had a case with the stem noct as in
Latin noctis, noctem, nocti, it would have to be-
come noeth in "Welsh, and this actually occurs in
trannoeth, ' the following day,' literally ' over^
night,' and in trannoeth the word noeth must he an
accusative, which is the case tra governed, as may
he learned from the fact that its Irish counterpart
tar always governs that case. Beunoeth, ' every
night,' is also an accusative, and so probably is
the 0. "Welsh form henoith (written henoid in the
Juvencus Codex), superseded later by heno ' to-
night,' which seems to be a shortened form of
he-nos: compare he-ddyw, ' to-day.'

So far of nominatives and accusatives : as to
the other cases, it is exceedingly hard to distin-



(delwedd B6154)


guish them from accusatives or from one another
now that their distinctive endings have been dis-
carded. We have, however, undoubted genitives
in ei, ' his,' ei, ' her,' and eu ' their,' which
have already been mentioned. The dative next:
years ago attention was called by Mr. Norris to
the pT/n in er-k/n, ' against,' as the dative of pen,
' head.' Now erbi/n is in Irish letter for letter
arckiunn, composed of the preposition ar and ciunn,
the dative of cenn, ' head : ' the latter is, however,
separable, admitting pronouns between the pre-
position and the noun, as in armochiunn " ante
faciem meam, coram me ; " and so the 0. Cornish
er y lyn would suggest that in Welsh also one
might at one time say er ei lyn, where we now
have to say yn ei erbyn, or Vw erbyn, ' against him,
to meet him.' Mr. Stokes has pointed out another
similar dative in 0. Cornish in such a phrase as
mar y lyrgh (= Welsh ar ei ol), ' after him :' the
nominative is leryk. Lastly, we have one certain
instance of an ablative, namely, that of pmy,
' who,' in the particle po, of the same origin as
Latin quo. You will notice also that the same
use is made of them in both languages in such
sentences as Po anhawddafy gwaith, mwyaf y clod
oH gyflawni, " quo difficilius, hoc prseclarius."

Now that we are hurriedly picking up, as it
were, a few fragments of the time-wrought wreck



(delwedd B6155)


of our inflections, you may expect a word about
the Welsh genders. I need not prove that Welsh
once had three genders, that is, not only the mas-
culine and the feminine, but also a neuter, of
which we have a familiar relic in the demonstra-
tive hjn^ as in hyn o Mysc, ' this much learning,'
hyn win, ' this much wine : ' add to this the 0.
Welsh pad = Lat. quod, quid. But more interest
attaches to the feminine : put together, for in-
stance, merch, ' a daughter,' and tlms, ' pretty,'
and they have to become merek dlos, ' a pretty
daughter.' Now, why is the t of the adjective
reduced into d? Well, if you remember what
was said on another occasion, it can only be be-
cause merch once ended in a vowel, and I hardly
need state -that that vowel was probably a or a.
Thus merch dlos represents an earlier merca tlos or
rather merca tlossa, for the a of the adjective is
even more certain than that of the noun, seeing
that it is to the influence of that a on the timbre
or quality of the vowel in the preceding syllable
we owe our having still two forms of the adjective,
tlws in the masculine and tlos in the feminine.
Tlws and tlos belong to a class of adjectives",
already noticed, which conform to the same rules,
and you may take the pair llym, mas. Hem, fern.
' sharp,' as typical of another, and as supplying
us with the principle which guides us in distin-



(delwedd B6156)


guishing the gender of monosylla'bic nouns : thug
if you propose to a monoglot Welshman any
monosyllabic nouns with which he is not familiar,
he will treat those with ro or y as masculines and
those with o or e as feminines, and in so doing he
thinks he is guided by instinct. This is probably
not the only habit of later growth which has been
mistaken for instinct ; and if you wish to find the
key to it, you have to trace it back in the language
to a time when the latter was on a level, so to
say, with Latin and Greek as regards the inflec-
tion of its substantives, while the origin of the
same habit must be sought thousands of years
earlier, when neither Celt nor Teuton, Greek nor
Roman, had as yet wandered westward from the
cradle of the Aryan race in the East.

Perhaps it is even more surprising to find in
later "Welsh traces of the dual number, seeing that
the very oldest specimens of its inflections which
the Aryan languages afi'ord us look weather-worn
and ready to disappear. But to give you an in-
stance or two in Welsh : we meet in the Mabinogi
of Branwen Verch Llyr with deu rcydel uonllmn,
that is, in our orthography, dau Wyddel fonllwm,
' two unshod Irishmen ' (Guest's Mabinogion, iii.
p. 98). Now in the singular we should have
Gnyddel bonllrmn, and in the plural Gmyddyl bon-
Uymion ; so it may be asked how it is that we have



(delwedd B6157


bonllwm made in our instance into fonllmm. There
is only one answer : Gwyddel must in the dual have
once ended in a vowel, and a glance at other
related languages which have the dual, such as 0.
Irish, Greek, and Sanskrit, would make it pro-
bahle that the vowel in question must have been
the ending of the nominative or accusative dual ;
but instead of guessing which the vowel or vowels
were in which the dual ended in Early Welsh,
perhaps the best thing would be to ask you to take
a look at that number in Greek in which our
instance might be literally rendered : hvo avviro-
Si]T(a ToiBe\e. Instances are not very rare in Med-
iaeval Welsh, but I will only mention one or two
more : in the Mabinogi of larlles y Ffynnawn
we meet with deu was penngrych wineu deledwiv:,
" two youths with beautiful curly hair " (Guest's
Mab., i. p. 35). A still more interesting instance
occurs in William's " Seini Greal," p. 91, where
we read of deu deirw burwynnyon, ' two pure- white
bulls.' In Modern Welsh there is one instance
which is well worth mentioning. The Carnarvon-
shire heights, called by English tourists " The
Rivals," have, from the Carnarvon side, the ap-
pearance of three peaks forming two angles or
forks between them : hence their Welsh name is
Yr Eifl, which has been supposed to be plural ; but
were it so, it would be, not Yr Eifl, but Y Geifl or



(delwedd B61


Y Gq/lau, the singular being ffo/l, ' the fork.' So
Tr Eifl means, I cannot help believing, the two
forks, and might be rendered into Greek Tw "Ayxr],
but that we should thereby lose the connotation of
the Welsh name, which in this instance, as ia
so many other Celtic place-names, turns mainly
on a metaphorical reference to the configuration
of the human body.

Interesting as the foregoing instances may be
to us, as persons whose language is the Welsh,
you must not suppose that they enhance materi-
ally the certainty with which glottologists regard
the former inflections of Welsh substantives ; for
they are satisfied that Welsh is near of kin to
Irish, and that Irish had the inflections in ques-
tion, not developed in the course of its own history,
but inherited from of old from an older language
which was the common mother of Irish and Welsh.
The discovery in Welsh of a few such remains
as have just been pointed out, they would have
thought uot improbable beforehand, but suppos-
ing, on the other hand, that that did not occur in a
single instance, they would not have felt in the least
dismayed. Where, then, seeing that Welsh still
shows traces of at least five cases, three genders,
and three members, does the improbability lie of
its having retained the endings indicative of some
of them — say the nominative and genitive singular



(delwedd B6159)


masculine — as late as the 7tli century ? Nowhere,
it seems to me. But as the transition of a lan-
guage from the inflectional to the positional stage
is an importalit one, which could not help register-
ing itself in its literature, let us turn our atten-
tion for a moment to this point. For our purpose
the difference between an inflectional and a posi-
tional construction admits of easy illustration. In
Latin there is no material difference of meaning
between rex Romm and Romce rex, that is, if we
put N. for nominative, and G. for genitive, both
sequences, N. G. and G. N. are admissible in that
language, while in Welsh we have to be contented
with N. G. only, and say brenhin JRhiifain, as
Rhufain hrenhin would not convey the same mean-
ing. Probably, however, when Welsh had case-
endings, it could have recourse to both N. G. and
G. N. ; but when the former were discarded one of
the latter had to be given up — that turned out to
be G. JH'. But the sequence JS^. G. could not
have beaten the other off the field in a day, and we
have to ascertain if any survivals of G. JST. occur
in the Welsh literature which has come down to
our time. A perusal of the poems attributed to
the early bards would convince you that such do
occur : I will only quote (in modern orthography)
a few at random from Skene's Four Ancient Books
of Wales : — cenedl nodded, " the nation's refuge ''



(delwedd B6160)


(ii. p. 7) ; huan heolydd wrfnAdd, " bold as the sun
in his courses " (ii. p. 20) ; Cymmerau trin, " the
conflict of Cj'mraerau " (ii. p. 24) ; rhiain garedd,
" delight of females " (ii. p. 93) ; and " Gorchan
Cynfelyn cylchwy nylad,^^ " Gorchan Cynfelyn, to
make the region weep " (ii. p. 96). Now, with
such survivals as these and others of a different
nature, which could be pointed out in the poems
alluded to, before our eyes, the conclusion would
seem natural that Welsh may well have retained
case-endings in common use as late as the 7th
century. On the other hand, it has, it is true,
been argued that the original composition of the
poems in question took place long before the 12th
century. But what concerns us here is the fact
that the evidence they give us, taken for what it
is worth, affords a presumption that one is right
in supposing case-endings to have been in use in
our language as late as the 7th century ; and the
outcome of all this is, that thus far we have not
met with any prima facie reason whatever for
thinking that the old Celtic monuments still ex-
isting in Wales were not intended to commemo-
rate persons who spoke our language, or a language
which has, by insensible degrees, grown to be that
which we speak.

Now we move on to meet those who claim some
of our inscriptions as belonging, not to the Welsh,



(delwedd B6161)

ńLECXUKlli 1> . lO-l

but to the Irish. You will find their views advo-
cated, though not without eliciting opposition, by
some of the writers who contribute to the Arclimo-
logia Cambrensis. It is by no means irrelevant to
our case that you should know that they are men
whose study is archaeology rather than the Celtic
languages. For though the belief in the Irish origin
of inscriptions found here may have originated in
the discovery that seme of them are written in
Ogam, a character once supposed to be exclusively
Irish, it now rests mainly on other arguments,
which can have no weight in the eyes of any one
who has enjoyed the advantage of a glottological
training. Thus, whenever an early inscribed stone
is discovered here bearing a name which happens
to be known to Irish annalists, it is at once as-
sumed that the inscription containing it is of Irish
origin. But this, it requires no very profound
knowledge of the Celtic languages to perceive, is
perfectly unwarrantable. For as Welsh and Irish
are kindred tongues, and as their vocabularies of
proper names of persons must, at one time, have
been identical, the occurrence of the same Celtic
names in Wales and Ireland is just what one is
entitled beforehand to expect. Neither, supposing
a name, to put the case still stronger, forming part
of an early inscription in Wales not to be trace- .
able in later Welsh, while it happens to occur in




(delwedd B6162)


Irish 1)00118, can the inscription be claimed as Irish :
besides, it would warrant our advancing similar
claims. For instance, we might say, If onr
stones with the name Decceti on them are Irish
because we have not as yet succeeded in tracing it
in Welsh books, whereas it is thought to be de-
tected in Irish ones, then on precisely the same
grounds we claim the Irish stone bearing the name
Cunacena until the latter can be shown to occur in
later Irish, as we have it in the successive forms
Cunacenni, Concenn, Cincenn, and Kyngen, this
side of St. George's Channel. The one claim is
as good as the other, and neither deserves a hear-
ing; for the question as to which Celtic names
have survived in Wales and in Ireland respec-
tively belongs to the chapter of accidents, and the
wonder, perhaps, is that the instances are so nu-
mferous as they are of the same ones having come
down to the Middle Ages or to modern times in
both countries.

If you were to press the advocates of the Irish
claim for their reasons, the answer would be of
the following type, which I copy from the Archceo-
logia Cambrensis for 1873, page 286 : " Were I
to find on the shores of Wexford or Waterford a
sepulchral inscription to Griffith, ap Owen, I should
be fully as justified in claiming it to be Irish as
Mr. Rhys is in claiming Maccui Decetti [szc] to



(delwedd B6163)


be Welsh." This is d propos of an Anglesey
inscription reading: Hie lacit Maccu Decceti.
Now this involves the fallacy of assuming that the
difference between Welsh and Irish has always
been so great as it is in modern times. If there
is anything I have especially endeavoured to im-
press on your minds in the previous lectures, it is
the fact that the further back we trace the two
languages, the more strongly are they found to
resemble one another. There is one word in par-
ticular which Irish archaeologists, with a turn for
what may not inappropriately be termed simple
inspection, have made a great deal of — I mean the
word maqvi, the genitive of the word for son.
This, it is said, is the Irish mace or mac, ' a son,'
genitive maicc or maic, and it is held to settle the
question. The truth, however, is that it contri-
butes nothing at all to the settling of it ; for, as
all Oeltists know, the Kymric languages syste-
matically change qv into p, so that the 0. Welsh
map, now mab, ' a son,' is as regularly derived in
Welsh from maqv-i as mac is in Irish. What
would have been to the point would be to prove
that the Kymric change of qv into p was obsolete
before the period of the inscriptions whose origin is
in question. This the writers whose views we
are discussing would, I feel confident, find to be
an impossible task to perform, and the attempt



(delwedd B6164)


would, moreover, be likely to take them out of the
beaten path of simple inspection, one of the most
recent outcomes of which may here be mentioned,
as it will answer the purpose of a reductio ad ah-
surdum of this way of appreciating old epitaphs.
In the churchyard at Llanfihangel y Traethau,
between Harlech and Portmadoc, there is a stone
bearing an inscription apparently of the 12th
century : one line of it reads Wleder matris Odeleu,
whence we find elicited totus, teres atque rotundus,,
the full-grown Irish name Dermot O'Daly : this,
3'ou will be surprised to learn, was not meant as a
joke — see the Archceologia Camhrensis for 1874,
page 335.

Though the reasoning which seems to have led
to the conclusion that our early inscriptions are
Irish will not for one moment bear examination,
that conclusion may, nevertheless, be the only
one warranted by the facts of the case ; hence it
is clear that we must not dismiss it until we have
considered how it deals with them. Well, the
first thing that strikes one here is the arbitrari-
ness of a theory which, from a number of inscrip-
tions, would select some as being Irish without pre-
dicating anything of the remaining ones, or assign-
ing the principle on which the selection is made.
You might perhaps expect that those written in
Ogam would be the only ones claimed as Irish,



(delwedd B6165)


and at one time it was so ; but eventually it was
found convenient to cross that line ; and no
wonder, for, as you must have noticed, there is
no essential difference between those partly
written in Ogam and those written in Eoman
letters exclusively. So Welsh antiquaries could
hardly have been taken by surprise by a sweeping
statement of the Irish claim, such as we meet
with. in the Arch. Camhrensis for 1873, p. 285,
in respect of the names Vinnemagli and Senemagli
in a Denbighshire inscription. There we read,
'•' Both of the names in question are Irish, as
are most, if not all, the names found on those
monuments hitherto known as Romano-British."
This you will keep in mind as a concession on
the part of our Irish friends of the fact that the
nanfes in our inscriptions are of a class, and do
not readily admit of being separated into such as
are Irish and such as are not.

Then, by supposing some of the epitaphs to be
commemorative of Irish pagans of a very early
date, they involve themselves in difficulties as to
the crosses to be frequently met with on them.
This, however, may be a mere instance of chrono-
logical extravagance not essential to the theory,
but it would not be so easy to take that view of
an assumption to which few would be found to
demur, namely, that the pagan Irish did not use



(delwedd B6166)


the Roman alphabet. We observe, therefore, with
some curiosity how they extricate themselves from
the difficulty arising from the fact that almost all
our inscriptions are partly or wholly in Roman
letters. As to those which are exclusively in the
latter, the oracles have not yet spoken ; at any
rate, I cannot find their utterances. But in the
case of stones bearing inscriptions in both charac-
ters, if the one is not a translation of the other,
then the Roman one owes its presence on it to a
Romanised Briton having seized on the monument
of a Gael to serve his own purposes, there being,
it would seem, a great scarcity of rude and un-
dressed stones in those days. If, on the other
hand, the one merely renders the other, the
explanation offered is somewhat different. The
following, which I copy from the Arch. Cambrensis
for 1869, p. 159, relates to the bilingual stone
at St. Dogmaels, near Cardigan, reading Sagrani
Fili Cunotami, and in Ogam Sax/ramni Maqvi
Cunatami : — "The story of the stone looks like
this ; that it was erected as a memorial over some
well-known chief of the invading Gaedhal, who for
a long period occupied South Wales, and that at
some period after, when thelanguage of the Gaedhal,
and the use of Ogham were dying out, some patriotic
descendant of the hero, to perpetuate the memorial,
re-cut the inscription in the Roman characters then



(delwedd B6167)


in use; the monument is of great antiquity, the
Eoman inscription alone, on the authority of Mr.
Westwood, being referable to a date ' not long after
the departure of the Romans.' " Ah uno disce omnes.
A still greater difficulty presents itself in the fre-
quent occurrence on the stones in question of names
which to most men would seem to be Latin, while
it is, on the other hand, acknowledged that the
Goidelic race was never conquered by the Romans,
and that they would otherwise have been too proud,
as we are told, to adopt Roman names. How this
difficulty is disposed of as a whole I do not know.
However, I find that Turpilli and Victor are made
out to be pure Irish ; but whether the same fate
awaits such names as Justi, Faternini, Paulini,.
Vitaliani, and the like, remains to be seen ; for the
possibilities of O'Reilly's dictionary of Modern
Irish are many. Unfortunately, such is the
reputation that work enjoys, and such are the
discoveries to which it helps men ignorant of
Old Irish, that an appeal to it on their part has
the charm of the last straw that broke the camel's

The foregoing are a few of the difficulties attend-
ing the claim made to our inscriptions. Now, I
would call your attention to particular instances of
them, which cannot, I think, be Irish : —

(1.) We will begin with a stone at Penmachno,



(delwedd B6168)


in Carnarvonshire, which reads : Cantiori Hie Jacit
Venedotis Give Fuit Consobrino Magli Magistrati.
Despite the waywardness of the Latin, it un-
doubtedly shows that the person commeniorated
was a man of importance, and a Venedotian citizen,
whatever that may exactly mean. ' The Venedotians
are not generally supposed to be of the Goidelic
race, and, as they are not likely to have made a
foreigner a citizen of their state, the conclusion is
unavoidable that the inscription is not of Irish
origin. It is much in the same way that one may
look at another which reads : Corbalengi Jacit Or dous.
The stone stands on an eminence overlooking the
Cardigan Bay, between the convenient landing-
places of Aberporth and Traethsaith, in Cardigan-
shire ; but I am inclined to think that Ordous
means that the person buried there was one of
the Ordovices of North . "Wales. If so, whether
he came there as an invader or as an ally, the
position of the stone, which seems to occupy its
original site, explains why it was thought expedient
to specify his tribe on his monument. So this also
could not well be Irish.

(2.) The inscription at Llangadwaladr, not far
from Aberffraw in Anglesey, reads Catamanus
Rex Sapientisimus Opinatisimus Omnium Regum.
It is right to state that it is not in Roman
capitals, but in what may be called early



(delwedd B6169)


Hiberno- Saxon characters, and that it is as-
cribed by archaeologists to the 7th centnry.
There are, however, other reasons for ranging it
with those of the Brit- Welsh, rather than with
later ones. It is probable that this Catamanus
was the Catmaa or Cadfan whom Welsh tradition
mentions as the father of Cadwallon and the
grandfather of Cadwalader, who is usually called
the last king of the Britons ; Cadwallon died,
according to the Annales Cambrice, in the year 631,
and the year 616 has been given by some Welsh
writers as the date of Cadfan's death. However
that may be, we are pretty safe in assigning it to
the 7th cetitury, and the inscription commemora-
tive of him dates, probably, not long after his
death. Whether Catamanus and his name are
likely to be claimed as Irish I do not know, but
the latter undoubtedly bears a family likeness to
several of those contained in our early inscriptions
so claimed. The same likeness is also observable
in the names of the kings of the Britons to
whom Gildas, writing not later than the middle
of the 6th century, undertook to give a good
scolding. They are the following, all except the
first in the vocative: — Constantinus (king of
Damnonia), Aureli, Vortipori (king of the Dime-
tians), Cuneglase (rendered by Gildas into Latin
as Lanio fulve), and Maglocune, supposed to be



(delwedd B6170)


Maelgwn, the king of Grwynedd, who, according
to the Annales Cambrice, died in the year 547.
Now these, as well as Catamanns, must be sur-
rendered as Irish, if our early inscriptions are
rightly claimed as such.

(3.) An instance, which has already supplied us
withauame of interest, occurs on a stone near Whit-
land, Carmarthenshire, which reads Qvenvendani
Fill Barcuni. Now in Irish genealogies one finds
the name Qvenvendani matched most exactly by a
Cenjinnan, to which a parallel is offered in the Four
Masters' Annals of Ireland in a name Ceandubkan.
These would be, in Mod. "Welsh, Penmynnan and
Pendduan, but as far as I know they do not occur.
However Penmynnan has its analogy in Carn-
Tcennan, ' Arthur's dagger ; ' but Cenjinnan is a
derivative from a still more common Irish name,
Cenfinn, which would be in Welsh Penwyn,
' Whitehead : ' it occurs more than once in the
Record of Carnarvon, and we read of a lorwerth
Tew ap y Penwyn in Edward the Third's time
{Arch. Cam. 1846, p. 397). The portion of
our Qvenvendani (shortened probably from Qven-
navendani) represented by Penwyn and Cen-
finn is Qvenvend-, which accordingly contains
curtailed forms of the words for head and white,
that is, gven- and vend-. The modern forms are,
Welsh pen, Ir. ceann, ' head,' and Welsh ywy«.



(delwedd B6171)


' white,' feminine gwen, Ir. Jinn. You will here
notice the change of i into e before a complex of
consonants in the Welsh vend-. The i would re-
main in Irish, as we see ixoxsxjinn and Ptolemy's
BovovivBa, that is Buwinda, ' the Boyne : ' so in the
case of Gaulish names such as Vindos and Vindo-
magus ( = Welsh Gwgnfa, as in Llanfihangel y'
Ngnynfa in Montgomeryshire ; Irish, Finnmhagh,
'the white or fair field'). This makes it probable
that not only Qvenvendani cannot be' Irish, but also
Vendoni, Vendumagli, Vendubari, and Vendesetli in
other inscriptions. Still more decisive is the evi-
dence of Barcuni, which, I have no doubt, is the
same name as the Irish Berchon in Ui-Berchon,
Anglicised into Ibercon, and meaning literally the
descendants of Bercon ; but it is now applied, as
frequently happens to such names in Ireland, to a
district in the county of Kilkenny. This informa-
tion I derive from the entry for the year 851 in the
Annals of Ireland. In a note the editor, 0' Dono-
van, observes, that within the district alluded to
there is a village known as Rosbercon, anciently
called Eos-Ua-mBerchon. Now the Ixish Berchon
may be the genitive of Berchu, involving the
word cu, ' dog,' genitive con. So. the nominative
corresponding to Barcuni, which itself stands pro-
bably for an older Barcunis, may have been Barcu.
Barcu and Barcuni would now be in Welsh, if they



(delwedd B6172)


ouly occurred, Berchi and Berchwn respectively.
If you compare with the Irish Berchon our Barcuni
or Berchwn, you will observe that there is a pho-
nological discrepancy between them ; for Ber-
chwn or Barcuni ought to be in Irish Bercon, and
not Berchon. In other words, the Irish Berchon
could not be derived from Barcuni, but from a
longer form, Baracuni. Here, then, we have a
difference between the two languages which makes
itself perceptible elsewhere in such instances as
Welsh gorphen, ' to finish,' for morqvenn, and
Mod. Ir. foirckeann (also Scotch Gaelic), ' end,
conclusion,' for woriqvenn or woreqvenn. This,
you see, makes it highly improbable that Barcuni
is Irish ; hence it would follow that here we have
an early inscription of Welsh origin, in which
the place of later jo is occupied by qv, which in the
case of maqvi has been made so much of by Irish

(4.) The next pair of instances bears on de-
clension : the text is supplied in part by a stone
at Trallong near Brecon — it reads Cunocenni
Filius Cunoceni Hie Jacit. Here you see that as
we have a nominative Cunocenni and a genitive
Cunocenni (for we may venture to supply the
omitted n), the name must be one the stem of
which may be regarded as ending in i. Now
glottology teaches us that in the common mother-



(delwedd B6173)


tongue of the Aryan nations /-stems ended in
the nominative in -is, and in the genitive in
-ajas. The latter was variously contracted in
the various languages derived from it : thus
Sanskrit nom. avis, ' a sheep,' gen, ave^ or
avyas, Grreek ttoXk, gen. ttoXjos or TroXeto?,
Lithuanian akis, ' eye,' gen. ak'is. In very early
Welsh and Irish, or in the language from which
both have branched, we may suppose the ending
of the genitive of this declension to have been jas
(with ^' =r y in yes) , but not perhaps to the exclu-
sion of the longer -ajas. The names, then, in our
inscription may be restored thus : nom. Cunacennis,
gen. Cunacennjas, of which the latter seems to
have undergone contraction into Cunacennis ; so
that when the language began to drop final Sj they
became nom. Cunacenni and gen. Cunacennl, a
distinction which may not have been lost at the
time when the inscription was cut on the Trallong
stone. Let us now turn to the other side, and see
what would become in Irish of a Goidelo-Kymric
genitive of the form Cunacennjas. Clearly, if we
are to be guided by the ordinary rules of Irish
phonology, the j would disappear, which would
give us Cunacennas, and when the s followed the
example of the j, the word would be found reduced
to Cunacenna, which actually occurs written Cuna-
cena on an Ogam-inscribed stone found at Dunloe,



(delwedd B6174)


in the county of Kerry. It is, however, right
that I should tell j'ou, that in some of the earliest
Irish inscriptions both the s and the J (written ?)
appear intact ; for instance, on a stone found at
Ballycrovane, in the county of Cork, reading
Maqvi Decceddas Ami Toranias — the word awi
means grandson, and becomes in Old Irish manu-
scripts due, or, with an inorganic k, hdue. Lest
you should thiuk that all this has been excogitated
to suit my views, those of you who read German
— and I hope that by and by their number will
be considerable — will find that Ebel and Stokes
inferred genitives of this declension in -ajas and
Jos for Early Irish in the first volume of Kuhn's
Beitrcege, published in 1854, and that, most
likely, without having heard of the inscription
alluded to above.

(5.) If it should seem to you that too much is
here built on a single word, there remains one or
two other instances which cannot be passed over.
On the Anglesey stone already noticed we meet
with Maccudeeceti, which one might venture to
write Maccu-decceti, as forming one name, although
consisting probably of a noun governed in the
genitive by another. Compare also Maccodecheti,
on a stone now at Tavistock, in Devonshire. That
Decceti and Decheti are in the genitive is certain,
but our "Welsh data could not enable us to ascer-



(delwedd B6175)


tain the declension to which they belong ; so we
have to resort to Irish inscriptions in which the.
name in question occurs. The following are re-
ported : Maqvi Decceddas Ami Toranias, already
mentioned ; Maqqvi Decedda, found in the parish
of Minard, co. Kerry, now in the Museum of the
Eoyal Irish Academy in Dublin ; Maqvi Decceda
Hadniconas, found at Ballintaggart, with six
others ; Maqviddeceda Maqvi Marin, found at Kil-
leen Cormac. Now Welsh Decceti and Irish
Deccedas taken together prove that we have here
to do with an J-stem ; so the genitives may be
restored to the forms — "Welsh Deccetjas, Irish
Decceddjas or Deccedjas, for Irish seems to have
hesitated between the provected ddj or d'J and the
non-provected dj. The forms which occur in the
two languages give us the three stages Deccedjas,
Deccedjas, and Deccetjas, which require some notice
before we proceed further. In Welsh I know of
no closer parallel to tj for dj than that of llj
(mostly reduced to II) in such words as arall,
' other,' Iv.araile, from a stem ar-alj-, to be com-
pared with Latin alius; oil, ' all,' Ir. uile, from
ol^-; pebyll, 'a tent,' now 'tents,' from Lat.
papilio, ' a butterfly, a tent : ' to these may per-
haps be added an instance from one of our early
inscriptions, namely, Turpilli, on a stone near
.Crickhowel in Brecknockshire. This, no doubt.



(delwedd B6176)


stands for an earlier Turpilji, once the pronuncia-
tion, Welsh or Eoman, or both, of Turpilii, the
genitive of the Roman name Turpilius: compare
also jilli for filii or rather jilji. The provection
would lead to the inference that Decceti was ac-
cented Deccdti, whence it is clear that Vitaliani
on another stone need not have followed suit. In
point of fact, it seems to have become Guitoliaun,
which occurs in a MS. of Nennius, where we read
of Guitmd fill Guitoliaun, as though it had been
Viialis fili Vitaliani. As to the Irish provection
into dd, we find a good parallel to it in the U-
declension, which is thought to have once ended
in the nom. in -us, and in the gen. in -awas or
-was. Thus Mr. Stokes, in the volume just re-
ferred to of Kuhn's Beitrcege, p. 450, traces two
Irish genitives, tairmchrutto, " transformationis,"
and crochta, " crucifixionis," to tarmicru^ejas and
cruca^ijas respectively : compare also such genitives
as Lugudeccas, Rettias, Anawlamattias, said to occur
on early inscribed stones in Ireland. "What has
been hinted as to the phonology of Decceti is a
mere conjecture, to which I would add another, and,
perhaps, a better — namely, that the Welsh and
the Irish forms, taken 'together, may be regarded
as pointing to the still earlier ones Dencendis,
genitive Dencendjas. In case this hits the mark,
the word is to be referred to a root dak or dank.



(delwedd B6177)


whence we have Greek ZeUvvjii, Lat. dicere, Ger-
man zeigen. But, not to take up any more of
your time with these .details, the outcome of
them, as far as we are here concerned, is that
Cunocenni, Decceti, and Decheti are Welsh, while
the Irish forms are Cunacena, Deceddas^ and the
like. Consequently the inscriptions in which
the former occur cannot be Irish. We are
now enabled to return with greater certainty to
Corhalengi, which being a nominative, is likely to
be of the J-declension. Hence it .would also follow
that Evolengi and Evolenggi are of that declension,
which cannot in Irish make i in the genitive, as
these do; so it is unnecessary to say that the
inscriptions containing them cannot be Irish. The
same observations would seem to apply to those in
which the names Vinnemagli, Senemagli, or Seno-
magli, occur in the genitive ; for that these forms
belong to the /-declension is suggested by the
fact that we have Brohomagli in the nominative
in an inscription reading Brohomagli Jam Ic Jacit
Et Uxor Ejus Caune. Add to the foregoing, that
although the Early Welsh base whence our cad,
' battle,' must have been caiu, of the ^/-declension,
we have the compounds Rieati nominative, and
Dunocati genitive, .while the Mod. Irish is iDonn-
chadh, genitive Donnchadha ; which makes it im-
possible that Dunocati could be Irish. This is




(delwedd B6178)


the way I would reason, if I felt certain that the
case-endings here in question are not mostly Latin
rather than Celtic. The more I scrutinise them,
the more I am inclined to treat them as Latin,
especially such genitives as Dunocati, and such
nominatives, as Corbagni and Ctmnoceni, for Cor-
bagnis and Cunoeennis.. But it is to be noticed
that this only makes our case against the Irish
claim still stronger, and that one has only to regret
that so many of the inscriptions are less valuable
than could be wished as materials for the history
of Welsh inflections. As the allusion to Cunocenni,
Corbagni, and Dunocati as Latinised nominatives
may appear scarcely intelligible to those who are
acquainted only with the Latin ordinarily taught
in our schools, it is right to explain, that from the
time of the Gracchi or thereabouts the ending is
appears not infrequently instead of ius; as, for
instance, in Anavis, Ccecilis, Clodis, Ragonis, and
the like. Further, it is a rule in our Early In-
scriptions to leave out s final: the same thing
frequently happened also in Roman ones, so that
such nominatives occur in the latter as Claudi,
Minuci, and Valeri. For more information on this
point, see the second edition of Corssen's great
work on Latin, i. pp. 289, 758; ii. p. 718; also
Eoby's Latin Grammar (London, 1871), i. p. 120.
(6.) Besides the numerous nominatives made to



(delwedd B6179)


end in our Early Inscriptions in the Latin termina-
tion us, and the possible Latinity of some or all' of
those in i, there is an instance or two where the
former appears as o for the old Latin nominative
ending os. One of these comes from Carnarvon-
shire, and reads : Al/iortus Eimetiaco Hie Jacet.
The other is at Cwm Grloyn, near Nevern in Pem-
brokeshire : it reads in Ogam Witaliani, and in
Eoman capitals Vitaliani Emereto, of which I can
make nothing but nominatives, the Welsh having
perhaps never stopped to consider whether there
existed such a Latin name as Vitalianius to be
transformed into Vitaliani. Emereto would be for
Emeretos, or, as it would appear in our dictionaries,
emeritus. Similarly we have consobrino for the
fuller nominative consobrinos in the inscription
already noticed as reading: Cantiori Hie Jacit
Venedotis Cive Fuit Consobrino Magli Magistrati.
(7.) To the foregoing it should be added that
feminines making their nominatives in e, such as
Caune, Tunccetaee, and the like, are also probably
indebted for that e to the usage of somewhat late
Latin, which, in its turn, is supposed to have
borrowed it from Greek. In the Roman inscrip-
tions of the time of the Empire the names of
Greek slaves and freedwomen appear in abund-
ance, such as Agapomene, Euehe, Theophile, and the
like : after them were modelled Cassiane, Juliane,



(delwedd B6180)


Sabiniane, written also with ae for e, whence even
suah genitives as dominaes, vernaes, annonaes, were;
formed. Nominatives of the kind in question were
also not unknown in Eoman Britain. I have come
across the following in Hiibner's collection already
alluded to: — Aurelia Eclectiane, Hermionae,
lavolena Monime, Julia Nundinae (in the mu-
seum at Caerleon), and Simplicia Proce. On the
question of Latin nominatives in e and genitives in
es or aes, see Corssen, i. p. 686, and Roby's Latin
Grammar, i. p. 12L It is hardly necessary to re-
peat that the Latinisation here pointed out is
incompatible with the Irish claim as it has hitherto
been put.

(8.) In Early Irish the Z7-declension made its
genitive singular in os, liable to be reduced to o ;
and in the Early Irish inscriptions, of which
accounts have been .published, amounting to 120
or more, not a single genitive in u occurs, while
those in os, o, appear in due proportion. In our
inscriptions, on the other band, the same genitive
is either o or u. So far, then, as one can judge
from this, our inscriptions containing the genitives
Nettasagru and Trenagusu cannot be Irish.

(9. ) Maccu -Decceti and Macco -Decheti have been
mentioned together, and it may appear strange that
one has cc and the other ch. The explanation is
simple enough : in the interval between their dates
the language may have begun to change cc into ch,



(delwedd B6181)


and probably also tt, pp, into th, ph. Here may
be mentioned the inscription already cited as
reading Brokomagli Jam Ic Jacit Et Uxor Ejus
Caune, which is in much the same style of later
letters as the Tavistock Stone with Decheti, There
is an apparent inconsistency in Macco- retaining
its cc unmodified ; but the cc in Macco- represents
an earlier ng or ngh, and it would be contrary to
rule if it passed into ch in Welsh. In Brokomagli
the h was undoubtedly sounded like our modern
ch ; for in 0. Welsh the name was Brochmail, later
Broehuail. The same remark applies to the h
in the epitaph reading Velvor Filia Broho, which
seems to be of the same date as the other two. In
Broho and Brohomagli the syllable broh, that is
broch, probably represents an earlier brocc, as in
Broccagni, a name said to have been read on a
stone at Capel Mair near Llandyssul, which has
since been effaced by a bucolic Vandal. Broccagni
is familiar in the form Brychan, and is precisely
the Irish Broccdn borne by the author of a hymn
in praise of St. Brigit contained in the Liber
Hymnorum in the library of Trinity College, Dublin.
But how does this bear on our argument ? Simply
in this way :" the change from cc into ch is unknown
in Irish, whence it is impossible that the inscrip-
tions containing Decheti, Brohomagli, and Broho
should be of Irish origin.



(delwedd B6182)


Now that the Irish claim has been shown to he
untenable, we might he asked to show how the
details of the inscriptions, in so far ^s they are
Celtic, fit into the history of "Welsh inflections ;
but this is rendered an impossible task by the
meagreness of our data. However, we have at
least one inscription which seems to belong to the
transition period preceding the total disuse of cases
by the Welsh: I allude to one of the stones at
Clydai, in Pembrokeshire, which reads in debased
capitals Etterni Fill Victor, and in Ogam Ettern

W[ic]tor. Here Victori (iox Victoris) is out

of the question, but the discarding of the case
termination was in this instance favoured by the
fact that the nominative was Victor, while the
genitive might be Victor. The inorganic doubling
of t in Etterni is a feature common to it and the
Old Welsh of the Capella Glosses. I cannot leave
this point without noticing in a few words the
fate of the vowel, more conveniently than correctly
called the ' connecting vowel,' as, for instance,
the in Dunocati, which has been completely lost
in its modern representative Dinyad, pronounced
Dir^gad. That the connecting vowel in compounds
was sometimes obscurely pronounced even in Early
Welsh is proved, as has already been pointed out,
by such pairs of instances as Cunotami and Cuna-
tami; but when did it altogether disappear ? In



(delwedd B6183)


the last-named instances it cannot have done so
until the t had begun to be softened towards o?,
otherwise we should have Cunatam-i, Cuntam
yielding Cynnhaf, whereas the modern form is
Cyndaf. Moreover, in a few instances, the
number of which could no doubt be increased by-
careful reading, the vowel comes down in manu-
script. The place known to Welsh tradition as
Catraeth is called by Bede Cat&racta; in the
Juvencus Codex, the Latin word frequens is ex-
plained by the Old Welsh word Ut'imaur, which,
were it still in use, would now be lUdfawr,
with Hid- as iu erlid, ' to pursue,' and might be
expected to have nearly the same meaning as
gosgorddfamr, ' possessed of a large retinue or
following : ' in Gaulish it occurs as the proper
name Litumara (G*luck, p. 120). In the oldest
MS. of the Annates Cambrice we have not only
Chtenedote to compare with the later Gmyndyd,
' North Wales,' but also a mention, under the
year 760, of Dunnagual filii Teudubr, more cor-
rectly Dumn-Agual or JDuvn&gual. Later he is
called Dyfnwal, a name which in Early Welsh
would have been Dumnoval-i or Dubnoval-i. In
the Saxon Chronicle, under the j'ear 1063, we
meet with Rhuddlan, called Rudelan, a spelling
which is supported by the Doomsday forms Rothz-
lanum, and, with the soft dental slurred over,



(delwedd B6184)


Ruelan.- Lastly, Giraldus Oambrensis writes
Eudkelan, Bledkericus (Bledri), Rodkericus
(Rhodri), Ythewal (Idwal), Landinegath (Llan-
dingad). I place no implicit faith in Giraldus'
spelling, but it seems certain that the connecting
vowel continued to be pronounced, however lightly,
for a long time after the Welsh had given up the
habit of representing it in writing, and that there
can have been no break in this respect between
the pronunciation of the Welsh of the Early In-
scriptions and that of the 9th century glosses.

This is also the place to call attention to the
fact that the ordinary formula of our Early Inscrip-
tions, such as Sagrani Fill Cunotami, came down
to later times. Thus, for instance, an elegy to
Geraint, the son of Erbin, in which the Welsh
poet, as an eye-witness, describes Geraint's deeds
of valour in the battle of Llongborth, is headed
Gereint Fil Erbin in the Black Book of Carmar-
then as published by Skene, ii. p. 37. This Geraint
is probably the Welsh king who, according to the
Saxon Chronicle, fought against Ine of Wessex in
the year 710.

Lastly, supposing, per impossibile, the foregoing
reasoning to be inconclusive, we still have a weighty
argument in the fact, for such it seems to be, that
the Kymric race has occupied Wales, Cornwall,
Devon, and other parts of England, from the time



(delwedd B6185)


of the Roman occupation to oar own day, excepting
in so far as their territory has been encroached upon
by the English nation and language. It follows,
then, that the onus probandi remains with the
advocates of the Irish claim, and that they are not
at liberty to attempt to prove any of our inscrip-
tions to be of Irish origin until they have made
out that the same cannot be explained as Welsh.
Let it first be shown that they cannot be Welsh,
then they will have a right to make them out to
be Irish if they can, and, logically speaking, not
before, as we have a priority of claim, which
stands whether they attribute the inscriptions to
Croidelic invaders, or regard them as proofs that
the Goidelic race occupied this country before the
Kymry. For, in either case, the knowledge of
letters may be presumed to- have reached the
former, whether in Ireland or in the more inacces-
sible parts of the west of Britain, through the
latter, who must have learned (if they had occa-
sion for it) from the Romans how to honour their
dead with inscribed tombstones. That the Kymry
should have taught this to the Gaels and so far
forgotten it themselves as to leave us no monu-
ments, while the Gaels are alleged to have left
so many, is incredible.

Allusion has just been made to a theory which
not only makes the Goidelic race the first Celtic



(delwedd B6186)


inhabitants of Wales, but tries to prove tbeir oc-
cupation of most of North Wales to have lasted
down to the 4th or the 5th century. As .it is
supposed that the Irish claim to our inscriptions
derives considerable support from this theory, it
is necessary to examine it briefly before we have
done with this question.

From .what has been said on the classification
of the Celts in a previous lecture, it is already
clear that the Goidelic Celts cannot be said to
have inhabited Wales before the Kymry, but it
will, nevertheless, be desirable to ascertain what
this theory has to recommend itself, especially as
it is put forth on excellent authority. In the
first place, it is founded, to a considerable extent,
on Welsh traditions which are supposed to refer
to the expulsion of Gaels from different parts of
Wales in the 6th century ; but the same tradi-
tions are admitted, be it noticed, to speak of them
invariably as invaders. However, it derives most
of its support from Welsh place-names, which are
supposed to commemorate the sojourn of the Gael
by their containing the word Gwyddel, ' an Irish-
man,' plural Gwyddyl or Groyddelod: such are
Gwyddelwern, Llan y Gwyddel, Forth y Gwyddel,
Twll y Gwyddel, and the like. But it is not at
all clear to me how any such names can go to
prove the priority of the Gael over the Kymro in '



(delwedd B6187)


"Wales. For a certain number of the places con-
cerned have surely received their names within
this or the last century, particularly on the coast
and- wherever Irish workmen have been employed.
A good many more, probably, of them date during
the long interval between the last century and the
end of the 12th. Then, if any of them date still
earlier, they may possibly be accounted for by the
various descents made on our coasts in the 10th,
11th, and 12th centuries by Irishmen or Irish
Danes, and by the return of Welsh exiles, such as
Gruflfudd ab Cynan and Rhys ab Tewdwr, at the
head of a following of Irishmen. If, perchance,
any of them are older than the 10th century, it
would be natural to trace them to Irish saints,
Irish traders, and Irish invaders who visited this
country ; but none of these last or of the fore-
going would help to prove that Wales was wrested
by the Welsh from the Gael. Then there are
other deductions to make from the list; for many,
probably the majority, of the names adduced have
nothing whatever to do with Irishmen, there
being another word, gnyddel, plural grcyddeli (for-
merly, perhaps, also gwyddyV), which is a deriva-
tive from gmydd, ' wood.' The identity of form
between it and the word for Irishman is only
accidental, as the Early Welsh form of gwyddel must
have begun with a «? or «?, while the initial of that



(delwedd B6188)


of Gwyddel was g, which is proved by the Old

Irish Gaedel, Goidel, Modern Irish Gaoidheal,

with a silent dh, which has led to the simplified

spelling Gael. The common noun gwyddel, which

is no longer in use, means a brake or bush, as in

one of Englynion y Clywed, which runs thus {lolo


" A glywaist ti chwedl yr Enid
Yn y gwyddel rhag ymlid ?
Drwg pechawd o'i hir erlid."

In Dr. Pughe's dictionary, under the word erdd,
this is rendered : " Hast thou heard the saying of
the woodlark in the brake avoiding pursuit? —
bad is sin from long following it." Under the word
gTuyddelawg _ he gives tir gwyddelawg as meaning
" land overrun with brambles," and he rightly
renders gwyddelwern " a moor or meadow over-
grown with bushes." In the same way no doubt
Gmyddelfynydd is to be explained. So in the bulk
of instances like Mynydd y Gwyddel, Gwaun y
Gwyddel, Gwern Gwyddel, Nant y Gwyddel, Pant y
Ghfoyddel, Twll y Gwyddel, and the like, the word
gwyddel may be surmised to have no reference to
Irishmen. The outcome of this is, that after mak-
ing the deductions here suggested from the list,
there can be few, if any, of the names in question
which could be alleged in support of an early occupa-
tion of Wales by the Gael. They would undoubtedly



(delwedd B6189)


fall far short of the number of those with Sais, ' an
Englishman,' plural Saeson, such as Hkyd y Sais,
Pont y Saeson, and the like, of which a friend has
sent me a list of thirty instances : by a parity of
reasoning, these ought to go some way to prove
the English to have occupied Wales before our

It is- needless to repeat, that even were one to
admit the Gaels to have been the early occupiers of
this country, it would by no means follow that our
inscriptions belong to them and not to the Welsh.
On the other hand, as it cannot have been so,
our priority of claim to them remains untouched.
Lastly, it would not be exactly reasoning in a
circle to call attention, in passing, to a fact which
has an important bearing on the question of the
classification of the Celtic nations, namely, that
the controversy as to the origin of our inscriptions
rests entirely on the close similarity between Early
Welsh and Early Irish. Had they been less like
one a.nother, and had the primeval difference be-
tween them not been altogether imaginary, it could
never have arisen.

So far nothing has been said of the. pre-
historic period mentioned in the scheme laid
before you of the chronology of the Welsh
language. What happened to it during that
period can only be inferred, not to say guessed.
It is, however, by no means probable that the



(delwedd B6190)


Celtic immigrants into these islands found them
without inhabitants, or that they arrived in suffi-
cient force to exterminate them. Consequently it
may be supposed that in the course of ages the
conquered races adopted the language of their
conquerors, but not without introducing some of
their own idioms. The question, then, is who
these prEe-Celtic islanders' were, and whether the
Celtic languages still have non-Aryan traits which
may be ascribed to their influence. In answer to
the first of these questions, it has been supposed
that the people whom the Celts found here must
have been of Iberian origin, and nearly akin to the
ancient inhabitants of Aquitania and the Basques
of modern times. In support of this may be
mentioned the testinaony of Tacitus in the
11th chapter of his Agrieola, where, in default of
other sources of information, he bases his state-
ments on the racial differences which betrayed
themselves in the personal appearance of the
British populations of his day. Among other
things, he there fixes on the Silures as being
Iberians. The whole chapter is worth reproducing
here. " Ceterum, Britanniam qui mortales initio
coluerint, indigense an advecti, ut inter barbaros,
parum compertum. Habitus corporum varii:
atque ex eo argumenta. Namque rutilte Cale-
doniam habitantium comse, magni artus, Ger-



(delwedd B6191)


manicam originem adseveraut. Silarum colorati
vultus et torti plerumque crines, et posita coBtra
Hispania, Iberos veteres trajecisse easque sedes
occupasse, fidem faciunt. Proximi Gallis et
similes sunt; seu durante originis vi, seu procur-
rentibus in diversa terris, positio coeli corporibus
habitum dedit. In universum tamen asstimanti,
Gallos vicinum solum occupasse, credibile est.
Eorum sacra deprehendas, superstltionum persua-
sione : sermo baud multum diversus, in depos-
cendis periculis eadem audacia, et, ubi advenere,
in detrectandis eadem formido. Plus tamen fero-
cise Britanni prseferunt, ut quos nondum longa pax
emollierit. Nam Gallos quoque in bellis floruisse
accepimus : mox segnitia cum otio intravit,
amissa virtute pariter ac libertate. Quod Britan-
uorum olim victis evenit : ceteri manent, quales
• Galli fuerunt."

Accordingly, some of the non-Ayran traits of
Welsh and Irish may be expected to admit of
being explained by means of Basque. Unfortu-
nately, however, that language is not found to
assist us much, as it is known only in a com-
paratively late form. So we turn to other
prse-Aryan languages still spoken in Europe,
namely, those of the Finnic groups. These last
show a number of remarkable points of similarity
with the Celtic languages. Hence it may be sup-



(delwedd B6192)


posed — and comparative craniology offers, I believe,
no difficulty — that the British Isles, before the
Celts came, were occupied by distinct races of
Iberian and Finnic origin respectively, or else, in
case it could be made out that Basque is related to
the Finnic tongues, by a homogeneous Ibero-Finnic
race forming the missing link, as the saying is,
between the Iberians and the Finns. That some
such a race' or races once inhabited all the west of
Europe is now pretty generally believed.

Proceeding on the supposition that p was foreign
to the idioms of the insular, or, as they had now
better be called to avoid confusion, the Goidelo-
Kymric Celts, one may by means of names con-
taining it point out certain localities in the British
Isles . occupied by tribes which were not of a
Goidelo-Kymric origin. These fall into two groups,
with which we may begin from the north-west and
the north-east respectively. Ptolemy, who lived in
the time of Adrian and Marcus Aurelius, and
wrote a geography, calls one of the islands be-
tween Scotland and Ireland Epidium, and the
Mull of Cantyre ^E-n-iBiov axpov, apparently from
the people, whom he calls Epidii, and locates airo
Toi) E-TTiBiov aKpov (»s Trpos avaToXw;. Further, he
gives a town of the Novantae the name Lucopibia:
it is supposed to have stood near Luce Bay, in
Wigtonshire. All these names together with



(delwedd B6193)


Mons Granpius may well be supposed to refer
to localities to which the unabsorbed remnants
of a prse-Celtic race may have been driven by
the Celts. In the next place, he mentions a
people in Ireland called the Manapii, and a town
called Manapia, supposed by some to be the site of
Dublin. As to this side of St. George's Channel,
he calls St. David's Head ^OKTairiTapov aKpov, and
the old name of St. David's seems to have been
Menapia, whence Menevia, Welsh' Mynym. Now
it is known that there were also Menapii on the
coast in the neighbourhood of the Ehine, but
although they were a maritime people, it is hardly
probable that they had sent out colonies to Ireland
and Pembrokeshire. So I conclude that these names
are vestiges of a non-Ayran people whom the Celts
found in possession on the Continent and in the
British Isles. Nor have I mentioned all, for it is
hard to believe that none of the following names
also is of the same origin : Welsh Manaw, ' the
Isle of Man,' which Pliny calls Monapia and
Ptolemy MomoiSa; Mona, Welsh lf<?ra, 'Anglesey ;'
the Menai Straits or Meneviacum Fretum ; Welsh
Mynwy, ' Monmouth,' on the Monnow, in the terri-
tory of the ancient Silures; and possibly also
Manau Gododin in the North, and Momonia,
Mumhain, or Munster in Ireland.
As the outposts of the other group may be men-




(delwedd B6194)


tioned the Corstopitum or Corstopilum of the
Itinerary of Antoninus, which is supposed to have
received its present form in the latter part of
the 3rd century : it is believed that the site is
that of Corbridge in Northumberland. The next
is Epiacum, mentioned by Ptolemy as a town
of the powerful tribe of the Brigantes : it is
identified by some with Hexham, by others with
Lanchester, and by others, with more probability,
with Ebchester. Whether these two places were
Giaulish or Teutonic it is not easy to say, for they
cannot be very far from the district where Tacitus
detected a Teutonic population; but whatever
settlements there may have been on the coast from
the Tweed to the Humber, the Brigantes are said
by Ptolemy to border on the North Sea. Proceed-
ing south, we come next to Petuaria, the town of
the Parisi, on or near the Humber : it has already
been surmised that this was a G-aulish position.
We now «ome to the Iceni in Norfolk, who had
a king whose name, according to Tacitus, was
Prasutagus. Next we have Ptolemy's ToUapis,
supposed to be Sheppey, and his Eutupice, identified
with Eichborough in Kent. More inland we meet
with a people whom he calls KaTvev^(Xavol ol kclI
Ka-Tj-eXdvoi, possessing the towns of Verulamium
or Old Verulam near St. Alban's, and Salinse,
which has been sought for in Bedfordshire



(delwedd B6195)


and South Lincolnsliire. More to the west and
north, we find in the Itinerary of Antoninus a
place bearing the distinctly Gaulish name Penno-
crucium in the territory of the Cornavii, who may,
therefore, be concluded to be Gauls : the site is
identified by some with Peukridge in Stafford-
shire, and by others with Stretton. Add to these
vestiges of the Gaul the fact that we have Gauls
in the Belgse, who counted among their towns
AquEe Salis or Bath, and in the Atrebatii located
between them and the Thames. Compare also
what Caesar says on this point in the 13th chapter
of his fifth book. From these indications it
seems to follow that rather more than one half of
what is now England belonged in Csesar's time to
tribes of Gaulish origin ; that is to say, all east
of the Trent, the "Warwickshire Avon, the Parret,
and the Dorsetshire Stour, excepting a Kymric
peninsula reaching as far as Malmesbury, and
widening perhaps towards the south to take in
Warehara in Dorsetshire, where, it is said, there
are inscriptions of Kymric origin. Against this
may be set the Cornavii, whose territory consisted
of a strip of land running from the Avon along
the east of the Severn and stretching to the mouth
of the Dee. If you want the assistance of a
map, turn to Mr. Freeman's Old English History
(London, 1873), where you will find one of



(delwedd B6196)


Britaia at the beginning of the 7th century.
According to that, the tract of country which the
English- then ruled over south of the Humber
coincided almost exactly with the boundary of the
Gaulish portion of Britain which has here just
been roughly defined. This apparent recognition
of Celtic landmarks by the later invaders is a fact
the historical and political significance of which I
leave to be weighed by others.

This view of th'e extent of Gaulish Britain,
which, it hardly need be said, is a mere theory,
derives some confirmation from the river-names of
England, which contains, for instance, important
rivers of the name of S>tour in Kent, Suffolk,
Dorset, Warwickshire, and Worcestershire. Simi-
larly we have others bearing the name of Ouse,
such as the Sussex Ouse, the Great Ouse, with its
tributary the Little Ouse, and the Yorkshire
Ouse which meets the Trent on the borders of
Lincolnshire. Lastly, we find a Stratford Avon,
a Bristol Avon, a Little Avon in Gloucestershire,
a Hampshire Avon flowing past Salisbury, and
an Avon: entering the sea near Lymington. But
these last rivers are supposed to bear an undoubted
Kymric name. It is, however, an easy matter to
show that it is not so. In the Itinerary of Anto-
ninus we seem to meet with Avon in the form of
Ahona; the Modern Welsh for a river is afon,



(delwedd B6197)


whicli we pronounce avon, and this stands for an
earlier abona or amona, whicli would in the course
of phonetic decay have to becorae our a/on. Now
it happens that it was probably not a rule of Welsh
phonology to change b or jre into v till about
the 8th century : so it remains that we should
suppose this softening to have taken place in
English, or in the language of the British Grauls,
whom the English found in possession of the
country drained by the Avons. Possibly another
and an earlier instance occurs in the vn, or, as
it is usually printed, un of such Gaulish names
as Cassivellaunus, Vercassivellaunus, Segovellauni,
Vellaunodunum, as well, perhaps, as Alaunus,
Genauni, Icaunus, Ligaunus, and the like.
Welsh tradition has, it is true, made Cassivel-
launus into Caswallawn, and Caswallon, which
naturally takes its place by the side of Cadwallon,
Idwallon, and Tudwallon ; but it is by no means
usual for early aun to make awn, on in Modern
Welsh, whence it is possible that only the mall
of the Weigh names just mentioned is to be
equated with the veil of such Gaulish ones as
Cassivellaunus, and that the terminations are
completely different. In that case Cadwallon and
Cassivellaunus should be considered as standing
for Catuvelldn- and Cassivellamn-, the latter con-
taining a vellamn- which I would identify with



(delwedd B6198)


Walamn-i, a name which occurs on an Irish tomh-
stone now in the British Museum ; two of its
edges read Maqvi Ercias and Maqvi Walamni:
we farther seem to have the Gaulish equivalent
in VALLAVNiTS ou a stone at Caerleon. It
is needless to add that mn remained intact both
in Early "Welsh — witness Sagranmi — and in Old
Welsh, as, for instance, in the Juvencus Codex
in the verb scamnhegint, " levant," from scamn,
now yscafn, ' light, not heavy.' The softening
of m into v is not the only instance of Gaulish
outstripping Welsh in the path of phonetic decay.
Another familiar one of a different order occurs in
the of petorritum for ua or mo, still represented
in full by wa in the Modern Welsh pedwar, ' four.'



(delwedd B6199)

( 199 ) ńLECTUEE V.

" Y mae Uythyraetli y Gymraeg yn fater lied ddyrys ; ao y tnae
Uawer o ysgrifenwyr, yn enwedig y rhai ieuainc, yn Uawer rhy
fyrbwyll a phenderfynol yn ei gylcli, ac yn dueddol i feddwl eu bod
yn ei amgyffred yn drwyadl, pan y maent hwytliau, yn rhy fynych,
heb gymmaint a deal! elfenau cyntaf y petli y maent yn eu hystyried
eu hunain yn athrawou ynddo." — Daniel Silvan Evans.

In this lecture it is proposed to give a brief sketch
of the fortunes of the Roman alphabet among the
Kymry, and to follow it through the successive
modifications which it has undergone among us
down to the present day. For the sake of not
breaking on the continuity of its history, what I
have to say respecting the Ogmic system will be
reserved for another occasion ; for the same reason
also I have thought it advisable to omit a number
of details, otherwise highly interesting, as well as
all reference to the improved methods of dealing
with pronunciations inculcated with so much suc-
cess by Mr. Ellis, Mr. Melville Bell, and Mr.

The Eoman capitals found in our Early Inscrip-
tions are A, B, 0, D, E, F, G, H, I, L, M, N, 0,
P, Q, E, S, T, V, X. As to their formation, they
are mostly more or less debased, as arch^ologists



(delwedd B6200)


term it: — As in Eoman inscriptions, the letter D
is to be found occasionally reversed with or without
prolonging the perpendicular, so as to give it the
look of our minuscule d ; N and S also occur re-
versed, and the I, when final, is frequently placed
in a horizontal position, but in the genitive fili
it forms now and then a short stroke tagged on
to the short bar of the F and the end of the L ;
these are, however, by no means the only instances
in which it is of a smaller formation, as in Roman
inscriptions, than the other letters. Ligatures
are not at all unusual ; on the other hand, abbre-
viations are rare in our inscriptions of the earliest
class, and in this they strongly contrast with
Roman ones, as in fact they might be expected to
do, seeing that they are the work of a people who
was, to say the least of it, less given to writing
than the Romans were. A general survey of our
ancient monuments would convince one that the
style of the letters used was subject to a steady
change, which by the end of the Brit-Welsh period
had reached such a point that they could no longer
be conveniently called Roman letters. Hence it
is that they are variously termed Anglo-Saxon, by
those who are familiar with the use made of them
in Old English, and Irish by others who are better
acquainted with the Irish language, which is to
this day written in them ; while of late it has



(delwedd B6201)


been usual to make a compromise between the
English and the Irish by manufacturing for them
the adjective Hiberno-Saxon. But all this tends
to conceal their real origin ; for though this style
of letters became naturalised among our neighbours
in Ireland and England, it was among the Kymry
that it was developed and invested with an in-
dividuality of its own. Under the circumstances,
we are entitled to speak of it as Kymric, and to
call the individual characters Kymric letters. The
following are the forms in which they appear in
printed Irish : <^bctiep5hilmTiop4p-ircux.
The change from the capitals of the Eoman
period to the corresponding characters used by the
Welsh in the 9th and 10th century of course
did not, as has already been suggested, happen in
a day, and our inscriptions supply us with most
of the intermediate steps. But I could not hope
to make this perfectly clear to you without the aid
of good drawings or photographs of the inscrip-
tions themselves ; a deficiency which has quite
recently been met by the publication of them in
an easily accessible form by Dr. Hiibner of Berlin,
in a work entitled " Inscriptiones Britannias
Christianas (Berlin and London, 1876). A still
more elaborate work on the same subject is pro-
mised by the English palaeographer. Professor
Westwood, under the auspices of the Cambrian



(delwedd B6202)


Archaeological Association. To ascertain the re-
lative dates of our inscriptions, that is to say, to
arrange them chronologically, is the one leading
problem to the solution of which all investigations
into Kymric epigraphy ought to contribute : a
first rude attempt at this might be based on the
style and form of the letters to which your atten-
tion has been called. Thus all our non-Ogmic
inscriptions down to the beginning of the 12th
century or thereabouts might be classed as follows :
(a) Those cut exclusively in Roman capitals ; {b)
those in which some of the letters are found to
assume the Kymric minuscule form ; and {c) those
which consist entirely of Kymric letters. How-
ever, another step in the same direction would
probably bring one to modify and correct, by
means of grammatical and historical indications,
this very rough classification, with some such a
result as to distribute (a) between the Roman
and the Brit- Welsh period, leaving (fi) entirely to
the Brit- Welsh period and (c) mostly to that of
Old Welsh.

The next place must be given to a short account
of the values of the characters which have been
thus far occupying us, and for the present it
will be convenient to treat the inscriptions of
the Roman and Brit-Welsh periods as though
they were all entirely written in Roman capitals.



(delwedd B6203)


unalloyed and undetased. Generally speaking,
the letters may also be regarded as having the
same values as in Latin ; but in a few instances
that statement requires to be explained or

H. In occasionally writing oc and ic for hoc and
hie, the Welsh seem to have only imitated the
Romans, who, as early as the time of Augustus,
sometimes pronounced the aspirate and sometimes
not ; later the confusion became still more com-
plete : see Corssen's work already alluded to, i. 107.
Some difficulty is offered by the occasional use of
h for the guttural spirant ch ; for not only is
the sound of h known to become ch in Welsh, and
vice versa, but it seems certain that in Broho and
Brohomagli, the letter h represents the ch of the later
Brochmail and Brochmel, a sound we find so written
in Decheti for an earlier Decceti. It had also pro-
bably the same value in Alhorttcs. But how came
the Welsh to write h for ch ? It is probable that
h represented both the aspirate and the guttural
spirant in Old English, and it might be said that
we owe this use of it in our inscriptions to early
English influence ; but even could it be allowed
that all the instances in question date after the
beginning of the 7th century, that would hardly
seem probable. We have, therefore, to fall backj
perhaps, on the fact proved by Corssen (i. 97-99),



(delwedd B6204)


that the old guttural spirant ch, which the Italian
nations began at a very early date to reduce to h,
lingered on a considerable time in the Latin lan-
guage, which, however, assigned it a very inferior
part, and took no trouble to distinguish it in writ-
ing from the aspirate ever encroaching upon it.
It is possible that h pronounced ch continued in
popular Latin even later than Corssen would have
admitted, and that it is to this pronunciation con-
tinuing in the country after it had been given up
by the more genteel rerum domini in the city of
Eome, that the often-cited words of Nigidius
Figulus, a contemporary of Cicero, originally
referred : " Rusticus fit sermo, si aspires per-
peram." However that may be, if the guttural
spirant continued in vulgar or rustic Latin down
to the time of Julius Agricola — and Italy is a
land where dialects have always thriven — it could
hardly fail to have reproduced itself in the pro-
vincial Latin of Britain, and this would explain
how our ancestors came to represent it in writing
by h, and not by ch^ in words belonging to their
own language. But in what words would the
latter be likely to give them occasion to use it
before the departure of the Romans ? Not in
such as Brohomagli, for here the spirant only
came in some time after as the continuator of cc ;
it was late, also, no doubt, that initial sw became



(delwedd B6205)

. LECTUEE V. 205

ho; whence we have now hw in S. Wales, and
dm in N. Wales. There remain two combina-
tions where they may have had it — namely, in
words where we now have eh or h corresponding
to Irish ss (also written s), mostly for an original
hs, as in Welsh dehav, (also decheu, and even
detheu), ' right, south ; ' 0. Ir. des ; it is to this
origin I would refer the spirant represented by k
in Alkortu. The other is where we have t/i, with
vowel compensation, answering to Irish cki, as in
Welsh taitk, 'a journey;' 0. Ir. teckt, 'to go;'
Welsh 7w/tk, ' eight ; ' 0. Ir. ockt. The original
of this was kt, which the Goidelo-Kymric Celts
seem to have modified into ckt, and that possibly
before their separation into Kymric and Goidelic
nations. However, after weighing all the. diffi-
culties which beset this question, I am inclined
to think that though our ancestors may possibly
have heard k pronounced as cA in a few Latin
words, the use of k for c/i by them in writing
their own language is to be traced to the influence
of the Ogam alphabet, the discussion of which
will give me an opportunity of returning to this

L. On the stone at Llanfihangel ar Arth, we
have Fivs clearly cut instead of filivs. This spell-
ing is, however, to be traced to a Latin source :
see Corssen's work already referred to, i. 228,



(delwedd B6206)


where such instances as _fiae foTjilim, Corneius for
Cornelius, and the like, are cited.

JS'c, Ng. On one stone we have Tunccetace and
on another Evolenggi, while the same name occurs
also as Evolengi. The digraphs nc, ng, were pro-
bably meant to represent the nasal gutturals, surd
and sonant respectively. Such forms as nuncquam,
conjuncx,juncxit, extincxit, and the like, occur in
Roman inscriptions of the time of the Empire.
Names in agn, such as Ereagni and Maglagni, ap-
pear later as Erehan and Maelan; so -agn must
haye passed into -angn towards the close of the
Brit- Welsh period, though the spelling in the in-
scriptions in point gives us no clue to the change :
later angn was simplified into an. Had the lan-
guage followed suit with the Irish, which has re-
duced -agn into -an, we should have had not
Erehan and Maelan, but Erehaen and Maelaen ;
possibly in some instances -angn may have yielded
-awn by a change of ng into w, which occasionally
occurs : see the Revue Celtique, ii. 192.

Np occurs, if I may trust my last attempt to
read the Cynffig stone, in the name Punpeius,
more commonly met with in books in the form
Pompeius. It was not unusual, Corssen (i. 263)
tells us, in Latin inscriptions of the 3d, 4th,
and 5th centuries, to write not only np, nb, but
also mt, md, the reason being, as he thinks, that



(delwedd B6207)


neither n nor m was clearly pronounced in such
positions : they seem to have served merely to give
a nasal effect to the vowel going before them, and
they were, accordingly, often left altogether un-
represented in writing. From 0. Latin Corssen
quotes as instances Poponi, Seproni, Noubris,
Decebris, and from late Latin cupare (= compare),
incoparabile, exeplu, Novebres: It is curious to
find that the epitaph just alluded to has Punpeius
rendered in Ogam hy a form beginning with Pope —
the rest of the word is now illegible, but it would
seem to have been Popei, for Pompei.

S. Final s is frequently omitted in our Early
Inscriptions, as, for instance, in the Latin words
cive, Ccelexti, Eternali, Nobili, Vitali, for cives,
Ccelextis, Eternalis, Nobilis, Vitalis. The same is
the case with nominatives singular of the second
declension when the vowel used is o, as in conso-
brino, Eimetiaco, Emereto, for consobrinos, Eimetia-
cos, Emeretos. But in case the vowel chosen was
the later u, the s is written as in Curcagnus,
Ordous, Saturninus, and even in Eoman inscrip-
tions nominatives in us and o are, as far as I can
ascertain, more numerous than those in u and os.
No nominatives in is for ius (see Corssen, i. 289,
758) retain their final s in our inscriptions, except-
ing Venedofis, which I take to mean Venedotius,
on one of the Penmachno stones. In popular



(delwedd B6208)


Latin final s probably dropped out of the pronun-
ciation at an early date, whence it naturally fol-
lowed that men who nevertheless had an idea that
some forms had a right to it, occasionally inserted
it in the wrong place : among other instances,
Corssen (i. 293) gives the genitives meis, Mercuris,
Saturnis, and the ablatives Antios, domus, junior es.
We seem to have an instance of the same kind in
the Ti'efgarn inscription, reading Nogtivis Fill

X. The combination xs for x is exceedingly
common in Roman inscriptions, and we meet with
it on the Trefarchog stone in the Latin word uxsor,
which, however, occurs written uxor on the Voelas
Hall stone. At a comparatively early date x, that
is cs, had got to be frequently pronounced ss or s,
whence a good deal of confusion between x and s
in writing. Such instances as vis for vix, visit
for vixit, and ye lis tor Jelix, are to be met with,
and vice versa one finds milex for miles, and xancto
for sancto (Corssen i. 297, 298). The only instance
of this kind which we have is Ccelexti, for Ccelestis,
on the Llanaber stone, near Barmouth. But that
the reduction of x into ss or s cannot have been
general in Latin before the Romans came in con-
tact with our ancestors, is proved by the fact of its
yielding in Welsh words borrowed from Latin, not
s simply, but s preceded by vowel compensation



(delwedd B6209)


in cases where a; followed close on the tone-vowel,

as for instance in the three words which follow :

coes, ' a leg,' from coxa, ' the hip,' llaes, ' slack,

long,' from laxus, and pais, . formerly pels, ' a

coat, a petticoat,' from pexa, that is pexa testis or

pexa tunica, though a somewhat different meaning

is usually ascribed to pexa in Horace's words,.

when he says : — •

" Si forte subucula pexse
Trita subest tunicse vel si toga dissidet impar,

J. A word, in the next place, as to the semi-
vowels j' and V. The Romans at one time used to
write eiis, Gaiius, peiius, Pompeiius, and to sound
them ej^us, Gajjus, pejjus, Pompejjus with _; (= y
in the English word yes or nearly so) ; but that
does not help us much with our inscriptional
forms Lovernii, Seniargii, and Ma..ani, where the
n can hardly have meant i or ij, but either _/z or iji.
Another curious case is that of mvliiek, for mulier,
on the Tregaron stone at Goodrich Court. Here
the second I may be due to thoughtlessness on the
inscriber's part, but I see no reason to think so.
It may be looked at another way : possibly it was
his intention to represent correctly his pronuncia-
tion of the Latin mulier as a trisyllable, so that
what he meant was mulljer ; but that is hardly
probable, as the inscription seems to be by no
means one of the earliest, and as it would have- been



(delwedd B6210)


more in accordance with the habit of our ancestors
to have treated mulier as muljer. So it remains
that we should regard the pronunciation intended
as being muljjer, and ihejj as a parallel to the ww
of Ilwweto written in Ogam on the Trallong stone,
near Brecon.

V. Latin v was probably pronounced like
English w, and the combination vu was frequently
reduced to u in the popular Latin of the time of
the Empire : among the instances given by Corssen,
i. 321, are aus,Jlaus, noum, for avus^Jlavus, novum.
We seem to have an instance of this on the Penbryn
stone in Ordous, which probably means Ordovus,
whence Ptolemy's plural OpSovtKe?. We have the
V doubled on the Glan Usk stone in pvteri for
pueri, and so in ntvinti at Cynwil Caio. They
are probably to be pronounced puweri and Nuwinti,
with the former of which compare povero men-
tioned by Corssen, i. 362, 668, as well as Italian
rovina as compared with ruina, and other cases
of the same kind. In Anglesey we meet with
ORVViTE, which may mean Oruwite or Ormwite.
If the preference be given to the latter, as I am
inclined to do, the spelling Orwite must be
regarded as dictated by the same cause as IlToweto
and muliier. Probably both jj and vv or rem
represent peculiarities of pronunciation which
cannot now be correctly guessed, and it is worth



(delwedd B6211)


Doticing that the semi-vowel in pvveei, orvvite,
and Ilwveto occupies just those positions where 0.
Welsh would give us ffu {—gw). So had we in-
stances of initial w or ww, nothing would be want-
ing to convince one that the digraph represented the
phonetic antecedent of our gu, gm. It is curious
to observe that pvveei has its exact parallel on one
of the few bilingual stones known in Ireland : I
allude to devvides on the Killeen Cormac stone
in the county of Kildare.

The doubling of consonants took place as in
Latin, especially where it was warranted by pro-
nunciation and etymology : this would be the case
in accented syllables. Even when the doubling
dictated by the etymology of the word was not
favoured by the presence of the accent, it seems
nevertheless to have been the rule, but it was
liable to be forgotten by the inscribers, as for
instance in Enabarri for Ennabarri, Fanoni related
to Fannuci, Qvenatauci for a name I should con-
sider more correctly written Qvennatauci, Tovisaci
for Tovissaci, and Trihni for Trilluni. Towards
the end of the Brit-Welsh period we meet with
opinatisimus and sapientisimus, and altogether s is
seldom doubled, but IVenegussi occurs so written,
while the Pgam gives it as Trenagusu. It is
possible that the nominative Cunocenni was
paroxytone, while its genitive Cunoceni was a



(delwedd B6212)


perispomenon ; but no ingenuity could discover
reasons for the spelling Vendubari as compared
with Barrivendi and Enabarri, nor can Sagrani be
defended except as a defective spelling of Sagranni,
the Ogmic form being indubitably Sagramni : the
reduction of mn to nn was familiar in Latin as
early as Cicero's time, as when cum nobis and etiam
nunc were pronounced cun nobis and etian nunc :
see Corssen, i. 265.

A. A word now as to the vowels : short a at
the end of the first part of a compound appears to
have acquired an obscure pronunciation. In
Ogam it is always written a, as in Cunatami,
Cunacenniwi, Nettasagru, Trenagusu; so also in
the Latin version of the names Catamanus, Corba-
lengi, Enaiarri, Qvenatauci, Trenacatus. Advan-
tage seems to have been taken of the obscurity of
the vowel in question to give the compounds some-
what more of the appearance of Latin formations;
so we find it written o and e, as in Cunocenni,
Cunotami, Evelengi, with which compare the Irish
Evacattos, of doubtful reading, it is true, Seno-
magli, Senemagli, and Trenegussi. The o of
Catotigirni, tholigh probably of the same obscure
sound, is of a different origin, standing as it seems
to do for an earlier u: similarly the e of Anatemori
possibly represents an earlier i or ja, if one is to
analyse the name, not into Ana-temori, but Anate-



(delwedd B6213)


mcri, with anate representing what is in Mod. Welsh
enaid, ' soul,' and to regard the compound as mean-
ing eneid-fawr, magnanimus, fieyaXo'^lrv^o?.

E. According to Corssen, i. 325, short e had
two sounds in early Latin ; one of them ap-
proached that of i as in the words fameliai,
Menervai, mereto, tempestatebus. This may be
seen, he thinks, from the fact that in the lan-
guage of the educated it passed later into i, while
that of the people retained the old sound. This
twofold value of Roman e explains to some extent
the hesitation which the early Welsh' display in
the spelling of such names as Catotigirni, Teger-
nomali, Tegernacus, from a word tigern-, now
teyrn, ' a lord or monarch,' all from tig-, now ty,
' a house ; ' compare, however, our Qvici and the
Qweci of an Irish epitaph. As to Emereto on the
Cwm Gloyn stone, it is not Emeritus changed hy
the Welsh into Emereto, but written by them as
they learned it from Eoman mouths. Similarly
does, which occurs more than once for civis in the
Roman inscriptions of Britain, proves that we
owe the e in dve, for cives, on the Penmachno
stone, to no caprice of the inscriber. And it can
hardly be doubted that it was from this country
that the same pronunciation of Latin found its
way into Ireland, where it appears on the Killeen
Cormac stone already alluded to. To pass by the



(delwedd B6214)


Ogam on it, which, according to .the last account
of it, kindly sent me by Dr. Samuel Ferguson of
the Eoyal Irish Academy, should he read Uwanos
Awi Ewacattos, the Latin version is ivvene
DRTViDES, for iwENES DEVViDES, to he construed
in the genitive as meaning Lapis Sepulcralis
Juvenis Druidis. Of Latin genitives in es for is
Mr. Stokes has found traces in Irish manuscripts;
he mentions os turtores for qs turturis, in an old
Irish commentary at Turin ; see Kuhn's Beitraege,
V. p. 365, and compare our Res patres for Ris
patris, to be noticed later.

0. As in the case of e, so also o had two sounds
in early Latin (Oorssen, i. 342). The one was a
clear o, the other approached u, and passed in the
dialect of the educated into u, while popular Latin
retained the older sound. Not to go further than
the Eoman inscriptions of Britain, as edited by
Dr. Hiibner in the volume already more than once
referred to, it may be noticed that the more
formal and carefully executed of them follow the
rule of literary Latin; but when we come to the
names of tradesmen as stamped on their wares,
the struggle between o and m reappears, as in the
following names, which are all in the nominative
case singular : Cocuro, also Cocurus, Dometos,
Julios, usually Julius, Malledo, also Malledu,
Malluro, also Mallurus, Mercios, and Viducos,



(delwedd B6215)

ńLECTURE T. 215'

also twice Viducus, whence it would seem that-
the fashion tended to the use of u when the s was
retained, and o when it was not. That this hesi-
tation hetween o and u was bequeathed by the
Komans to their Kymrlc pupils is certain: witness
the following instances — consobrino for conso-
brinus, Emereto for Emeritus, servatur and amator
on the same stone ; and Punpeius for Pumpeius,
in ordinary letters, accompanied by Pope- for
Pompe-, in Ogam, on another stone. In the
same way as consobrino and Emereto, I would
also treat the early Kymric names Eimetiaco, in
ALHORTVSEiMETiAco, OQ the Llanaclhaiarn stone,
and Cavo, in cavoseniaegii, on the stone in
Llanfor Church, near Bala. This, unfortunately,
does not materially help us in deciding whether
the vowel which is written u and o in maccu and
macco, and in genitives of the U declension, such
as Trenagusu, was long or short, as an inter-
change of 5 with u is not out of the question.

A. Where we have aw in Mod. Welsh, the lan-
guage had at an earlier stage a with a pronuncia-
tion to be compared probably with that of a in
the English word hall or am in draw. This
would be the sort of vowel to occasion some hesi-
tation, in writing, between a and o. We have
it, accordingly, written a in Eimetiaco, Senacus,
Tovisaci, Tegernacus, Veracius, and Ł> in Cone-



(delwedd B6216)


toci and Anatemori, where mor-i is perhaps the
prototype of our marm' ' great,' while the a appears
unchanged in Cimarus on one of the Caerleon
stones of the Roman period, and invites comparison
with such names as Indutiomarus, Segomarus, and
the like. The same sound it is perhaps that meets
us in Daari, the syllable daar in this name being
probably of the same origin as the Greek Sa>pov, ' a
gift:' compare JtoSa)/jo9, 'HXi,oSa>po<;, 'AiroXKoBeopo?,
and the like. The doubling of the vowel was an
early expedient used by the Romans when they
wished to indicate thatitwas to be pronounced long,
but no trace of it appears in the Roman inscriptions
of this country. However, it is an expedient
which might suggest itself to anybody, and besides
in Daari we have it in a name beginning with
Cuur in an epitaph of a considerably later date on
a stone now in Llangaffo Church in Anglesey :
the same method of indicating long vowels was
also sometimes adopted by the Irish. It would
not be safe to compare Lovernii, Seniargii, and
the like.

E. The confusion of cs with S and even e was
common in late Latin : we have a good instance
of this in one of our inscriptions in the words
Servatur Fidcei Fatrie\_que\ Amator. Your atten-
tion was called in another lecture to the pro-
bability of feminine nominatives in e owing that



(delwedd B6217)


ending to a Latinising tendency. The most
trustworthy instances occur in the following in-
scriptions : —

1. Tunccetace Uxsor Daari Hie Jacit,

2. Evali Fili Dencui Cuniovende Mater Ejus.

3. Hie In Tumulo Jacit E...stece Filia Patef-
nini Ani xiii In Pa.

4. Brohomagli Jam Ic Jacit Et Uxor Ejus

5. Culidori Jacit Et Orvvite Mulier Secundi.

Besides these we have a fragment reading
Adiune; and another stone, the reading of which '
is extremely difficult, seems to yield us the
feminine nominative Cunaide. Then there re-
main two names in e' which it would be hazardous
to regard as feminine. The one is a genitive
occurring on the Llanwinio stone, which I read,
with considerable hesitation, Bladi Fili Bodibeve.
Here, if one treat Bodibeve as . a feminine, the
anomaly of the mother being mentioned instead
of the father has to be accounted for : so there
seems to be no alternative but to suppose Bodibeve
to be the father's name. The other instance is
Nogtene in Ogam, and accompanied in Eoman
capitals by Nogtivis Fili Demeti on the Trefgarn
stone. There seem^ to be no reason to expect a
Latinised form written in Ogam, so that Nogtene



(delwedd B6218)


•would appear to be, not a feminine nominative,
but a genitive like Bodibeve. If so, the final e in
both is perhaps to be regarded as a by-form of
the i of the genitive of the /-declension, just as
we have o and u in that of the Ł/-declension.
Here it should be mentioned that we have at least
one Early Welsh name containing e which later
yielded oe : I allude to Vennisetli on the Llansaint
stone — the name occurs later as Gnynhoedl and
Gwennoedyl, which, teach us that our hoedl, ' life,
lifetime,' was in Early Welsh setl-.


V. Early Welsh u must have had at least two
sounds, that of long u in Italian, German, and
English in such words as rule, food, and another
sound resembling French u, or our modern u=^ ii,
or perhaps intermediate between them; but this
will require some explanation. Many languages
have shown a steady tendency to let u (and some-
times m) gradually pass into i. Physiologically
speaking, this seems to mean that the pitch of the
resonance chamber formed by the mouth in pro-
nouncing u is gradually raised by shortening the
mass of air extending from the vocal chords to
the lips, in order to let them settle nearer their
position of rest, and reduce the tension of the
muscles called into action when the mouth has to
be maintained at its greatest length, as measured
from the vocal chords to the lips. When u passed



(delwedd B6219)


into i no break is likely to have happened in the
transition ; it will, nevertheless, be convenient to
fix on one or two intermediate stages correspond-
ing to the sound of French u or Greek v, which
nearly resembled French u and will here be used
for it, and our Mod. Welsh u, which comes near
German ii, which may here represent it. We
have thus the series u, v, ii, i, or perhaps better
still, u, 0, V, ii, i. As instances may be men-
tioned the following : Aryan au had been reduced
into '§, sounded like French u, in 0. English, and
by the 13th century it had so closely approached i
as to be confounded with it in writing. Or take
the case of Greek, in which <tv, for instance,
Doric TV, ' thou,' stands for tuam, as may be seen
from the Sanskrit form which is tvam ; but in
Mod. Greek the vowel v is further narrowed so as
to be pronounced now like i, excepting in the
Spartan dialect, where the old sound still seems
to be usual, a characteristic which the Greek who
pointed it out to me considered modern and vul-
gar ! In the same way Latin io has regularly
yielded its much narrower French representative,
and in German the sound written ic is to French-
men's thinking frequently pronounced i. Lastly,
Early Welsh 5 or m has given us our modern u
(= u), which is mostly pronounced i in South
Wales : this may be most readily exemplified in



(delwedd B6220)


the case of words borrowed from Latin, such as
durus, ' hard,' and labor, laidris, ' labour,' which
have given us our dur, ' steel,' and Uafur, ' labour,
tillage,' pronounced in S. "Wales dir and llajir
respectively. Curiously enough the same process
had gone on in Welsh at an earlier stage in its
history, namely in those words where Mod. Welsh
has i corresponding to Irish u: it was complete
about the end of the Brit- Welsh period, as hardly
a trace of the older vowel is to be met witb later.
This vowel perhaps never represented an Aryan
long u, but an u which became long in the course
of phonetic decay, as for instance in the case of
Mod. Welsh ci, '.a dog,' Irish cu, which stands
for a nominative cuans, as may be seen from the
cognate forms Greek Kvtav, Sanskrit qvd, Eng.
hound: so in Welsh ti, Irish tu, Lat. tu, Greek
axj, Sanskrit tvam, Eng. thou ; and so in another
group of words, which must here be mentioned at
somewhat greater length, namely Welsh din,
dinas, ' a fort, a town or city,' Irish dun, 0. Eng.
tun, Mod. Eng. town, which point to a Celto-
Teutonic base duan of the same origin, perhaps,
in spite of the aspirate, as the Sanskrit verb
dhvan, ' to cover one's self, to shut' There can
hardly be any doubt as to the identity of our
modern Dingad with Dunocati on the stone in
Glan Usk Park, whence it is highly probable that



(delwedd B6221)


the u in that name was .sounded towards the close
of the Brit- Welsh period more like our i than
our w. The change, however, in the direction of
i would seem to have commenced after the time of
Ptolemy the geographer, who gives the prototype
of our din, Irish dun, the form Sovvov (with Greek
ov = Latin u, or English u in rule), and that
whether the names in point reached him from
Wales, Ireland, or Gaul : witness the following —
from Wales, MaptSovvov, our modern Caerfyrdd-
in, ' Carmarthen ; ' from Ireland, the name of
a town which he gives as Aovvov ; and from Gaul,
AvyovardBovvov, Aovy^ovvov, Ov^eXKoBovvov, and the
like, all of which end in Latin in dunum. The
two Welsh series of u passing into i were not'
confounded, because they were not contempora-
neous, as will be seen on comparing our tud, for-
merly tut, Ir. tuath, ' a people or nation,' with
Gaulish names such as Toutissicnos, Toutiorix
(Welsh Tutri), and the Gaulish word toutius,
supposed to mean * a citizen or one of a tribe,'
and found written toovtiov<!, where Greek ov, as
standing for the sound of Latin u, made it neces-.
sary to write oov to represent the Gaulish diph-
thong ou ; it is very probable that Gaulish ou was
represented by ou or ou or some nearly related
diphthong also in the 'common language of the
Goidelo-Kymric Celts before their separation.



(delwedd B6222)


Eouglily speaking, then, the two series stood thus
as far as concerns their relative dates : —

Goidelo-Kymric. Early Welsh. Old Welsh. ModernWelsh.
U V or u i i.

Ou u OT V OX u u and i.

We have possihly a trace of the old spelling of
Bingad in Dwncat, in the lolo MSS., p. 96, but
better attested is Gurcu for the name otherwise
written Gurci. Whether the u in Dencui, Dinui,
and Sagranui is of the kind here discussed, it
will be impossible to say until one or more
of these names have been identified in a later

Ai. We have no satisfactory instances of this
diphthong; for Vailathi and Genaius, both from
Cornwall, are somewhat late and highly obscure.
Besides these, Cornwall offers us a name of far
greater antiquity on the stone at Hayle, which I am
inclined to read Cunaide; but others have been in
the habit of reading it Cunaido or Cunatdo in the
masculine. Supposing Cunatdo to be improbable,
we should in Cunaide or Cunaido have a compound
of the pretty familiar cun- of our early names, and
of the word which appears later in Welsh in the
form of udd, explained in Dr. Davies's diction-
ary as meaning dominus : it would seem to be
matched in Irish by the old name Oed-a (genitive).



(delwedd B6223)


later Aedh, Aodh, Eaodh, Anglicised Hugh, and
the late Mr. Stephens of Merthyr Tydfil was pro-
bably right in regarding the Aedd of Mod. Welsh
tradition as a Goidelic importation from North
Britain — see the Arch. Cambrensis for 1872, p.
193. If, then, Cunaide (or Cunaido) is the correct
reading we have here an instance of ai before it
was reduced to u.

Au. It is probable that this diphthong in Early
Welsh, or at least towards the close of that period,
consisted of a plus the sound of the narrow u al-
ready described, which would not be very far from
our modern au. The reason why I think so is
that I fancy that I find it later only as ei and ai.
The cases in point are Caune, Cavo, Qvenafauci,
Vedomaui, and Mauoh... To begin with Caune,
it can hardly be doubted that this is the name
which later appears in the form of Cein, now Gain,
and as an ordinary adjective cain, ' fair, beautiful,'
of the same origin as Gothic skauns, Ger, schon
' beautiful, handsome, fair,' — our ceinach, ' a
hare,' is not related, its cein- being, as pointed
out by Mr. Stokes, the continuator of ca{s)in,
of the same origin as Sanskrit gaga, 0. Prussian
sasin-, Ger. hase, Mod. Eog. hare. Our next in-
stance Cauo can hardly but be the prototype of
the well-known Welsh name Cei, later Cai, which
possibly comes from the same source as Cain. It



(delwedd B6224)


is right, however, to add that Welsh tradition
mentions a Cau or Cam, but he is generally men-
tioned as coming from Prydyn in the North.
Against this should he balanced the facts that,
while Welsh hagiology mentions only one Cau or
Caw, we find allusions to at least three persons of
the name of Cei or Cai, that Cai yields the deri-
vative names Caiaw or Caio, and Caian the name
of one of Caw's many sons who settled in Wales,
and that not many miles from Llanfor Church,
wherein is the stone bearing the name Cavo, is the
site of Caergai or Cai's Fort. So it seems that
the Welsh was Cei or Cai, while Cau or Cam not
only comes from the North, but also represents,
not Cavo or Cavus, but a name which in its
Latin form is found given as Caunus. Qvena-
tauci has not been identified, but the leading
element in the name is already familiar to you,
and tauc-i is possibly to be equated with Teic-an,
a name which occurs in the Liber Landavensis,
p. 201. Similarly in the case of Vedomaui and
Mauoh.., it is probable that mau-i and mau-o
are of the same origin as mai in Gwalchmai, and
we seem to have them .in the name Mei and its
derivative Meic in the same collection, pp. 199,
221, 260, 261. In Latin words the sound of au
was difi"erent, as that makes in Welsh successively
ou, eu, au, as in Welsh aur ' gold ' from aurum.



(delwedd B6225)


and Foul, Feul, Paul from Paulus—i\i& natu-
ralised Paul, with, u^ii, has been expelled in
Mod. Welsh in favour of Paul pronounced Pol, an
attempt to imitate the English : the Paulinus of
our inscriptions should yield in Mod. Welsh Peulin,
but I am not aware that it occurs, but we have a
Welsh derivative from Paulus, and that is PeulaUy
as in Llanbmlan, the name of a church in Anglesey.
It is to be regretted that Carausius is n-ot to be traced
in any later form known to Welsh literature.

EL We find ei in Eimetiaco, and its occur-
rence in Punpeius seems to indicate that it was
sounded not very differently from ei in Mod.
Welsh. Provisionally Alhortus Eimetiaco may be
rendered Alhortus ^re-hastatus, the Early Welsh
ei being the equivalent of Latin <es, genitive ceris.
In 0. Welsh we seem to trace it in the name
Ejudon, probably for Ei-judon, on a stone in the
neighbourhood of Llandeilo in Carmarthenshire ;
and it is probably the same name, in a still shorter
form, that meets us in the Mabinogion, ii. 206,
as Eidon, which was then probably pronounced
Eiddon. Further we have the same ei taking
the form ei and ai in haiarn, ' iron,' keiarnaidd,
' like iron,' However, I could not now enter
into the details of the history of these forms,
as they would take up more of your time than the
importance of the single vocable Eimetiaco could




(delwedd B6226)


justly claim in this lecture (see the remarks on the
Welsh names of metals at the end of the volume).
If now we review the ground which we have
just travelled over, everything seems to indicate
that, although the polite Latin of Eoman litera-
ture made its way, no doubt, into the families
of natives of rank in this country, the ground it
gained here was very inconsiderable as compared
with the conquests made by the Humble and
motky dialect of the legions of imperial Eome,
and those who followed in their train. This kind
of vernacular, so far as we know it from the marks
of potters and other tradesmen, may be said, both
as regards language and lettering, to pass imper-
ceptibly into the Latinity of our inscriptions of
the Brit- Welsh period. Consequently those who
try to estimate the date of the latter by the ex-
tent to which they have been debased, in point of
language or lettering, as compared with the com-
paratively faultless official inscriptions emanating
from the Eoman army and its officers, cannot help
incurring the risk of dating the Brit- Welsh ones
all too late. For it is not an unusual thing to
find that a debased letter, for instance, which
does not appear in official inscriptions, was, never-
theless, in common use among the tradesmen of
the time, Had Tacitus had to write of the later
history of the Eoman occupation, he would pro-



(delwedd B6227)


bably have given more room to questions of .lan-
guage than he does in his account of Agricola's
successful policy, when he says in the twenty-first
chapter of that work : " Jam vero principum filios
liberalihus artibus erudire, et ingenia Britan-
norum studiis Gallorum anteferre, ut, qui mode
linguam Eomanam abnuebant, eloquentiam con-
cupiscerent. Inde etiam habitus nostri honor et
frequens toga. Paullatimque discessum ad deli-
nimenta vitiorum, porticus et balnea et convivio-
rum elegantiam. Idque apud imperitos humanitas
vocabatur, cum pars servitutis esset."

Another point worthy of notice here is the fact
that our inscriptions seem to prove, beyond all
doubt, that Latin continued to be one of the lan-
guages used by our ancestors for a long time after
the departure of the Eomans, and after the British
Church had acquired strength enough to secure it
against speedy extinction. Eventually no doubt
the vernacular of the Eoman tradesman passed
into a kind of ecclesiastical Latin ; but from the
1st century to the 10th its history in the west of
Britain probably knew no entire break, and Bede's
words cannot perhaps be quite irrelevant, when he
says that the island was in his time, the earlier
part of the 8th century, divided between five peo-
ples, the English, the Britons, Scots, Picts, and
Latins. This brings us down to the 0. Welsh period.



(delwedd B6228)


The alphabet in use in the specimens of Old
Welsh extant consisted of the following letters in
their Kymric form : a, b, c, d, e, f, g, h, i, 1, m,
n, 0, p, r, 8, t, u. X occurs in Nemnivus's alphar
bet; 5 and }) only occasionally appear, and m is
to be met with only in proper names in Asser's
Latin writings.

B. The leading value of this letter was no
doubt the same which we still assign it. But
the Eomans began as early as the 2nd century to
write b for v, and from the beginning of the 4th
century on their archives are said to show in-
stances of this in abundance : witness such forms
as Flabio for Flavia, Balentiniano for Valentiniano,
Nerha for Nerva, and salbus for salvus. This
habit of course found its way among the Welsh,
hence we find properabit for properavit on a cross
at Margam, and lob in the Ovid Glosses for what
was later written lou, now Jau, 'Jove.' But the
use of b for v by the Kymry in 0. Welsh and in
Latin must have been far more common than
these two instances would suggest, otherwise it is
difficult to see how it could have been regularly
adopted in 0. Irish in such words as fedb, Welsh
gweddw, ' a widow ; ' tarb, Welsh tarw, ' a bull ; '
serbe, Welsh chwerwedd, ' bitterness.' The confu-
sion of b and v in writing makes it very hard to
ascertain when b began to be reduced to v in



(delwedd B6229)


Welsli pronunciation. That such a reduction had
beguij very early in the 0. Welsh period is ren-
dered probable by the fact, that the labial is
occasionally elided in our earliest specimen of
manuscript Welsh, the Capella Glosses, as for in-
stance in tu, ' side,' for tub, tuv, 0. Ir. toib, and in
luird, i.e. luirth, ' gardens,' for lubgirtk, the plural
of a word now written lluarth, Mod. Ir. lubhghort.
C has never had the sound of s in Welsh.
Ch mostly had its present value of a guttural
spirant : occasionally it is found written he, and
sometimes the h is not written at all. It is to be
noticed that once it is written for gh, namely, in
inhelcha, " in venando," in the Capella Glosses ;
but it does not follow that it was then pronounced
as gh, it being possible that gh had been dialecti-
cally provected in pronunciation into ch in this

D, d, t, th, dd, 8, J). The chief use of d in 0.
Welsh was np doubt to represent the same sound
as in Modern Welsh, Besides that, it had also
to stand for the consonant we now write dd and
Englishmen th (as in this), but probably only where
that consonant had taken the place of an original
_;. At any rate we have no indication that d
began to be reduced into this sonant spirant until
towards the close of the period. In one instance
the Welsh borrowed the 0. English «? with a stroke



(delwedd B6230)


through the stem (S) to represent the sound of
our dd, namely, in the Lichfield Codex in }n ois
oisou^ " in sseculum saeculorum," — this is now yn
oes oesoedd. Mr. Stokes identifies our llawenydd,
'joy,' 0. Welsh leguenid with 0. Ir. Idine, and
suggests as a possibly related word the Lavinia
of Eoman legend, all of the Ja-declension : so -id
in the following stanza, which occurs in the
Juvencus Codex, stands for zS : —

" Na mereit mi nep leguenid — henoid
Is discnir mi coueidid
Dou nam riceus unguetid."

Further, as d could represent our sonant spirant
dd, for which we may also use S, it came, by a
little sacrifice of accuracy, to be occasionally used
for the corresponding surd th, as in luird, for luirth,
and papedpinnac, for papetkpinnac, ' whatsoever,'
in the Capella Glosses. This confusion points to
English, in which the uncertainty as to the use of
d, S, th, and J) has given rise to much discussion.
The last mentioned character, a D with the
stem prolonged both waysj was also occasionally
borrowed by the Welsh to do duty for the digraph
th, as in joej) in the Juvencus Glosses, and once
in the Oxford Cornish Codex we find S used for
th in lai^-mer, Mod. Welsh llaeth, 'milk.' Now
as (^ = S could do duty for th, so vice versa, th
could be used for a? = S, and further, as th was



(delwedd B6231)


used by some as a mere equivalent for t — more
strictly speaking it meant an aspirated t, as in 0.
Welsh hanther, 'half,' from a manuscript which also
shows jomjo^e^, 'fifth' — especially in writing Latin,
we find t also occasionally standing for the spirants
th and 8, as for instance in the Ovid gloss gurt, for
jurth, ' against;' and in the tract on weights and
measures in the earlier Oxford Codex we have
both petguared part and petguared pard for pet-
guare'6 parth, now pedwerydd parth, ' fourth part ; '
but still more interesting is the marginal gloss in
the Juvencus Codex, which is read issit padiu itaw
gulat, and should be treated as iss iS pad iu i'Sau
gulat, meaning literally, est id quod est illi patria :
the words meant to be explained form the relative
clause in the following : —

" Cunctis genitoris gloria vestri,
Laudetiir, celsi thronus est cui regia caeli."

But elsewhere in the sanae manuscript we have

irkinn issid crist, ' what Christ is,' with d for S.

Accordingly the Welsh stanza just- mentioned

would be a little more accurately written thus : —

Na mereit mi nep legueniS — iienoith.

Is discnir mi coueithi^ ,
Dou nam liceus unguefcS. ,

The habit alluded to of treating t and th as
equivalents is plentifully illustrated by Giraldus
Cambrensis in the way he transcribes Welsh



(delwedd B6232)


names such as his Thaph or Tapfi, ' the river
Taf,' Llandinegath for Llandinegat, ' Llandingad,'
Rothericus for Eotericus, ' Ehodri,' and the like ;
but he was so far impartial that he occasionally
also wrote ck for c as in Gueneloch, ' Wenlock,' and
Oscka, ' the Usk : ' similarly Uicemarch in his life
of St. David writes Theibi for Teibi, now Teifi,^
' the river Teivi.' The same habit is conspicuous
in the Cornish Vocabulary printed at the end of
the Grammatica Cdtica. We trace it still earlier
in Contkigirni, now ' Cyndeyrn,' in the oldest
manuscript of the Annates CaTnbrice, the writer of
which more frequently, however, asserts the equi-
valence of th and ^by writing t and c for the spirants
th and ch, as in Artmail and Brocmail for Arthmail
and Brochmail. The latter is also written without
h, as is likewise Eutychius, in Bede's Historia
Ecclesiastica, where, on the other hand, we have •
Meilochon, a form of the name Maglocunus inter-
mediate between it as used by Gildas and our
modern Maelgwn — in fact the person referred to
by Bede is called by Irish annalists Maelcon (see
Keeves' edition of Adam-man's Life of St. Columba,
pp. 148, 371). Add to this Cluith and Alcluith,
which Bede so writes for Cluit and Alcluit. In
all these instances and the like, ck, tk, pk were
either aspirated c, t, p, as in brick-house, pent-
house, and uphold, or simple c, t, p.



(delwedd B6233)


F would seem to have had the same soimd in

0. Welsh as our jf now. It occurs mostly in words
horrowed from Latin,' and as the initial of Welsh
words which originally must have begun with sp :
take for instance ^er, 'the ankle,' Greek a^vpov,

Jraetk, 'eloquent, loquacious,' Ger. sprechen, 0.
Eng. spr^can, now speak.

G had the value of our modern g, which is
never that of Eng. J. It had besides that of the
corresponding spirant, as heard in some of the
dialects of North Germany in such words as sagen,
lage, and the like : possibly also that sometimes
heard in the German words liegen, degen, and the
like. To avoid mistakes I should further specify
that the sounds I mean are those technically written
j/^ and y respectively by the German phonologist
Briicke and his followers, and g^ andj by Sievers
in the Bihliothek Indogermanischer Orammatiken
(Leipsic, 1876). That g between vowels or after

1, r had been pretty generally reduced to a spirant
in 0. Welsh is rendered highly probable by the
fact, that later it disappeared altogether in those
positions, and that in the oldest manuscript Welsh
it is sometimes written and sometimes omitted.
Thus we have telu (for teglu), now teulu, •' a family,'
as well as nerthheint, "armant," by the side of
scamnhegint, " levant," all three in the Juvencus
Codex ; and te (in dolte), now tai, ' houses,' in-



(delwedd B6234)


stead of teg, the plural of tig, now ty, ' a house/ in
the Capella Glosses, among which we meet also
with paulloraur, a kind of collective plural ex-
plaining pugillarem paginam, and appearing with-
out the g of the h&tm pugillares, ' writing- tablets.'
But in this last case it would perhaps be more
correct to suppose that a y or g/i (=y^ —3^)
has become u just as we have had to point out in-
stances of another g ov gk (=g^=j) becoming^ in
such words as arjdn and Morjen : for more in-
stances of u for g see the Revue Celtique, ii. 193,
iii. 87. Gh is actually once found so written in
Ovid's ArtofLove, namely, mhelghati, "venare,"for
helgha ti, now helja 'di, hela di, or hel di, ' do thou
hunt.' Mention has already been made of the
spelling helcha, to which a kind of parsiUel is offered
by the Latinised form Pembroehia, whence pro-
bably the English Pembroke : the 0. Welsh must
have been Penbrog or Penbrogk, which is now, of
course, Penfro.

H. This was, no doubt, the representative of
.the aspirate in 0. Welsh as it is in Mod. Welsh ;
but was it also used for ch in 0. Welsh? We
meet certainly with the words hui and suh, of which,
however, the latter is Cornish, as it comes from
the later Oxford Glosses : in the Juvencus Codex
it is duly spelled such, " vomis," and as Cornish
was in the habit later of eliding h—ch, it is not



(delwedd B6235)


at all certain that it was intended to pronounce suh
as if it had been -written such. Then as to hui, the
probability is that in 0. "Welsh it was pronounced
with A, and that the latter has since been pro-
vected into ch, as the word is now chwi, ' you.'
The reason for such a change would be the pneu-
matic pressure alluded to in connection with initial
ffh passing into ch. But chmi, for 0. Welsh hui,
is exceptional in that it belongs to all Wales,
while in most other instances cAw is confined to
N." Wales, and km holds its ground' in S, Wales,
Eeturning, then, to the use of h as the exponent
of the aspirate in 0. Welsh, I may here cite a re-
mark made by Mr. Ellis in his work on Early
English Pronunciation, ii. p. 598 — it is to the
following effect: "Uneducated speakers, espe-
cially when nervous, and anxious not to leave out
an h, or when emphatic, introduce a marked k in
places where it is not acknowledged in writing or
in educated speech." Now this, especially the
allusion to emphasis, although written with re-
gard to the treatment of k in English, calls
attention to a principle which has played a part of
some importance in the formation of words in our
own language, seeing that it loves to aspirate the
accented vowel in the middle of a word, as for in-
stance indihdreb, ' eb^voyevh,' diarhdbol, 'proverbial.'
Some, it is true, wish to ignore this k in writing.



(delwedd B6236)


and believe it to be the outcome of a modern cor-
ruption ; but that seems to be a mistake, for
cuinhaunt, ' deflebunt/ nerthheint, " armant/'scawM-
hegint, " levant/' are as old as the Juvencus Codex,
and nobody perhaps would now object to glanhau, ' to
cleanse,' eyfjawnhau, ' to justify,' although the h in
them also is merely the accessory of the stress-ac-
cent, while such words as coffdu, ' to call to memory,
are altogether left out of the reckoning, although
their ^ only stands for an earlier y>^, so that coffdu
represents eqfhdu. The case is the same where the
accent has since retreated, as when we have coffa
instead of coffdu, or lloffa, ' to pick up with the
hand, to glean,' for llof-hd, from llof—Uaw, ' hand,'
as in llofrudd, also Uawrudd, ' a murderer,' literally
' red-handed.' Still older, perhaps, is the case of
pedol, ' a horseshoe,' from the Latin pedalis, ' a
slipper,' which appears in the Welsh of the 12th
century as pedhaul, that is, ped-hdul, whence later
petaul and pedol. By the side of pedol may be
^l&c&A. paradmys, 'paradise,' which in that case
cannot, be derived from -n-apaBeiao?, but from a
Latin paradlsus, if the latter may be supposed to
have been pronounced paradeisus by those from
whom the Welsh borrowed the word. But for the
h evolved by the accent, we should now have not
pedol and paradnys, but peddol and paraddwys.
And it is as the accompaniment of the stress-accent



(delwedd B6237)


that I would regard the aspirate in the following
words : — Casulheticc, "penulata," in the Capella
Glosses, where we have also ellesheticion, "mela,"
where the writer had perhaps at first intended only
to write elleshetic, and afterwards added a syl-
lable on finding that mela was plural — at any rate
that this enigmatic word was accented ellesMticion
is in the highest degree improbable. The Juvencus
Codex has crummanhuo, " scropibus," ceroenhou,
. " dolea " (which suggests that plurals in ou were
formerly oxy tones), and apassive T^\\i-ra\planthonnor,
" fodientur," as well as the cuinkaunt, nerthheint,
scamnhegint already mentioned. Among the Ovid
Glosses we have guorunhetic, " arguto." The
later Oxford Codex (Cornish) offers us brachaut
{=^brac-hdut') as well as Irracaut, " mulsum," and
Ainkam, ' oldest.' The effects of the same ac-
centuation is, perhaps, to be traced in the y of its
Mod. "Welsh equivalent hynaf, as well as in the
surd mutes of the degrees of such adjectives as
teg, ' fair : ' at any rate, until a better explanation
offers itself, I would regard teced, ' as fair,' tecacA,
' fairer,' tecaf, ' fairest,' as standing for teg-kddr,
teg-hdch, teg-hdf, though the latter do not occur,
and the former are only known in Mod. "Welsh as
paroxytones. It is in the same way, no doubt,
forms of the so-called future perfect should be
analysed, such as gwypo, (' that he) may know,'



(delwedd B6238)


and bythoch or bothoch, in books byddoch, (' that
you) may be.' Among 0. Welsh words which have
never been very satisfactory explained, and some
of which may contain an k of the origin here indi-
cated, may be mentioned anbitkaul, bemhed, dig-
uormechis, nemhe, roenkol.

In late Latin it was not unusual to write Ihesu
for lesu, eontroversikis for controversiis, and the
like. The same expedient was adopted in the
Cornish Glosses in such forms as bakell, "securis"
(but laubael, 'a hand-hatchet'), later Cornish
boell, Mod. "Welsh bwyall, ' an axe ; ' deleAid,
' a door-fastening,' Welsh ' dylaith ; guillihim,
" forceps," Welsh gnellaif, ' shears ; ' and gurehic,
' a woman,' Welsh gToraig. In instances of this
class the h was probably quiescent, but its use
was by no means confined to 0. Cornish, for we
find immotiMou, " gesticulationes," in the Capella
Glosses, and Jutkahelo (elsewhere Judhail, Ithael
ItheV) on a cross at Llantwit Major in Glamorgan :
the same abuse _ of the letter h is also abundantly
illustrated in the Venedotian versions of the Laws
of Wales. And now we may attack some of the
Breton forms in the Eutychius Glosses, such as
mergidhaham, '' evanesco." Here the first k seems
to be the accompaniment of the accent, while the
second looks as if it had been intended to stand
between the two as after the elision of the g,



(delwedd B6239)


which mtist have belongetJ to the word in an
earlier form mergidhagam, with which one may
compare the 0. Welsh scamnhegint, " levant,"
later yscafnheynt ; or else the pronunciation in-
tended was mergidhdm, with a long and, perhaps,
jerked or perispomenon. The other instances in
the manuscript in question appear with only one
of the two /j's : thus etncoilhaam, " auspicio aus-
pex," lemhaam, " acuo," but datolakam, ' I select.'
"With a few reservations, already indicated, one
may say that the best collections of 0. Welsh
words, namely, the glosses on Martianus Capella
and those in the Juvencus Codex, are on the
Tvhole accurate as far as conce'rns the letter k :
the latter, it is true, shows h once misplaced in
hirunn, for irhunn, now yr Awn, ' who,' and once
omitted in anter for hanter, ' half.* But the
writers of the glosses in the other codices, be-
sides indulging in an occasional heitham (for
eitham, now eithaf, 'utmost'), which seems to
point to the Grwentian dialect of parts of Glamor-
ganshire and Monmouthshire where no h is now
pronounced by the uneducated, either in Welsh or
English, unless it be in the wrong place, show
a decided objection to beginning certain particles
with vowels : thus they write mostly, but not
exclusively, ha for the expletive a before verbs ; ha,
hac, for a, ac, ' and, with ' — the h is still written



(delwedd B6240)


in Breton ; hai for a'z, ' and his ; ' ham for a^m^
' and my ; ' hi for i, '' his, her ; ' hin for in, nowyw,
' in ; ' ho for o, ' from ; ' hor for dr, ' from the.'
How they arrived at the idea of adorning these
monosyllables with an h, a habit which extended
itself even more indiscriminately in O. Irish, I
cannot guess, unless it was the result of being
used to write h, after it had ceased to be heard, in
the frequently-recurring Latin words hie, hcec, hoc,
and the forms immediately connected with them.

1. This letter stood in 0. "Welsh as in Mod.
Welsh both for the vowel i and the semi-vowel,
which, for the sake of distinction, is here written
_;'. In one instance, damcirehineat, " demorator,"
in the Capella Glosses, we have eat substituted, in
Old English fashion, for iat, that is, jaf. At any
rate there is no reason to think that the termina-
tion in question formed two syllables then any
more than its modern representative jad does in
our own day. One cannot be certain that the e
in the Latin word dolea, for dolia, in the Juvencus
Codex, is due to the same influence, for dolea is
known to occur elsewhere ; but no doubt attaches
to Margeteud for Margetjud, now Meredudd, on
the Carew Cross in Pembrokeshire.

L, II. 0, Welsh I had probably the same sound
which it has still, but in the former it is pro-
bable that it admitted of being aspirated when



(delwedd B6241)


it occurred as an initial or in contact with a pre-
ceding n and, possibly, r : at any rate, that seems
to have been the case in 0. Cornish, and I am
inclined to think 0. Welsh followed suit, though
it is the equivalent of II, and not Ih, that we seem
to have in the Capella gloss mellhionou, "violas,",
Mod. Welsh meilljon, ' clover, trefoil.' In 0.
Cornish It had become lit, and the t had been
assimilated, as proved by such forms as celleell
from cukellus, Mod. Welsh cyllell, ' a knife,' with
which compare the French couteau : similarly 0.
Cornish elin, " novacula," stands for ellin, Mod.
Welsh ellyn, ' a razor,' Irish alfan. . But besides
these 0. Cornish had initial M as in hloimol,
" glomerarium," and we have probably the same
hi or Ih in ehnlinn, which I take to mean enhlhinn :
the Mod. Welsh is enllyn, already alluded to. If
0. Welsh as well as 0. Cornish had both U and Ih,
then it follows that II has since extended its
domain in Welsh at the expense of Ih, which is
unknown in the language now, excepting perhaps
when yn mho, le, ' in quo loco ? where ? ' is dialec-
tically cut down into ymhli ? mhle ? or hie ? which
is also liable to become lie. That the spirant surd
which we write II existed in 0. Welsh, has been
shown in a former lecture ; but it is probable that
it was confined to words in which it represented
earlier l-l, or where it preceded t. In the latter




(delwedd B6242)


combination it was perhaps always written It, as
that could not lead to any confusion, and as lit
wanted etymological support : I can recall only one
instance in point in 0. Welsh, guogaltou, "fulcris,"
which occurs in the Capella Glosses. But con-
fusion might arise if II and I between vowels or
at the end of a word were not distinguished in
writing ; accordingly our authorities are as a rule
accurate in this respect, with the exception of the
Oxford Cornish Glosses, where about one-third of
the instances lack an I each, and that of the stanzas
beginning with Niguorcosam in the Juvencus
Codex : in them no consonant is doubled. Thus
they offer us ealmir for callaur, nouel for nouell,
patel for patell, and, to rhyme with the latter, a
conjectural canel for canell, possibly of the same
origin as the French cannelle, ' cinnamon : ' irre-
spective of this the number of the loan-words in
these stanzas is remarkabl-e.

M had probably the same value as at present.
In one instance, da.uu, " cliens," in the Ovid
Glosses, it seems to have been reduced to v, that
is dauu is to be read dauv, possibly with a nasal
twang imparted, as in Breton and Irish, to the
vowel by the m before it passed into v; but,
whether or no, the nasal is lost to Mod. "Welsh.
The modern forms of the word are daw, ' a son-
in - law,' plural dawon, but also dawf, plural



(delwedd B6243)


dojjon, which is not to be confounded with dofjon
the plural of dof, ' tame ; ' for the latter implies an
earlier dam-, Aryan dam-, while daw, dawf stands
for dam- of the same origin as the Sanskrit forms
ddmd, -ddma, ddman, ' a band, bond, fetter, tie.'
This enables one to account for what would now
appear a curious use of the word daw, in Brut y
Tyrcysogion (London, 1860), p. 118, where we
meet with the words y daw gan y chwaer, or, as we
now write, ei ddaw gan ei chwaer, ' his connection
by his sister,' that is in other words ' his brother-
in-law : ' compare the Ger. schnur, ' a cord, twine,
tie,' and schnur, ' a daughter-in-law,' which glot-
tologists, it is true, are in the habit of regarding,
for reasons not very evident to me, as in no way
connected. So much of the word dam: my
account of its origin in Kuhn's Beitraege, vii. p.
231, is utterly wrong. Whether the u of 0.
Welsh arm or enu, now enw, 'a name,' was arrived
at by reducing m into a nasal vowel, or by an
exceptional substitution of w for m, is by no means
clear : the Irish forms corresponding to 0. Welsh
anu, plural enuein are anm, plural anmann.

Ng, in 0. Welsh, as in Mod. Welsh, represented
the guttural nasal. The digraph got this value
all the more firmly attached to it when, in the
course of phonetic decay, nd, mb became nn, mm,
and lyg or ng-g in the same way lost its mute.



(delwedd B6244)


Previously the guttural nasal was mostly repre-
sented by the n in ng^ and so it continued in nc.
You will remember, however, our meeting with
Evolenggi and Tunccei;ace in surveying the previous
period. As a matter of writing the n is not
always found expressed at all in 0. Welsh :
thus we meet with cibracma in an entry in the
Lichfield Gospel for cibrancma, which probably
meant ' a place of battle,' from cibrane, now
cgfranc, ' a battle ; ' and in the Cornish Q-losses
we have torcigel, " ventris lora," for torcingel.
This would seem to have originated in the habit
of saving trouble in writing by omitting one or
more letters in a word, and indicating the place
of the omission by a touch of the pen above the
line : of course the latter was not infrequently
forgotten by careless writers, and, in the case of
"Welsh ng, this became, perhaps on the whole, the
custom ; for when original g non-initial regularly
disappeared, and when c had as yet not been com-
monly reduced to g, no great confusion could arise
from writing g for ng. It is thus that g is also
to be read in the Luxembourg Folio, which shows
no ng at all, in the words drog, " factionem,"
mogou, " comas,'.' rogedou, " orgiis," igueltiocion,
"in fenosa." Drag also occurs there written
drogn, where the influence is visible of gn, pro-
nounced ngn in late Latin in such words as mag-



(delwedd B6245)


nus, signum, and the like ; in fact, we have signo
written singno on the cross on Caldy Island.
But as to the habit of writing g for ng, it was
once so common, that one or two words of learned
borrowing from Latin must have been permanently
misread : I allude to the Latin Jlagellum, which
the Welsh treated as Jlangellum, and thence de-
rived the modern forms fflangell, ' a scourge or
whip ; ' another of the same kind was legio,
treated as lengio, whence our Biblical lleng, ' a
legion.' This was, of course, impossible in the
familiar name Castra legionum, which duly be-
came Caerlleon, ' Chester, Caerleon ; ' we have also
places called Carreg y Lleon and Hafod y Lleon in
the- neighbourhood of Bettws y Coed.

Ph had the same sound as at present, but it
seems to have been rarely used, f being preferred.
In a few instances p is written for ph, as in the
name Gripiud, for Griphjud, now Gruffudd,
' Griffith,' in the Lichfield Gospel.

jR had no doubt the sounds of our r a,nd of our
rA initial or following n, and the habit of writing
rh as if it were simply r will explain the spelling
of Hir-hoidl, as Hiroidil in the Gwnnws inscrip-
tion, which must be reckoned as belonging to this
period. The earliest written evidence to the exist-
ence of initial rh is perhaps the name Hris in the
Saxon Chronicle (in a manuscript marked Cott.



(delwedd B6246)


Tiber. B. i. in the Master of the Eolls' edition)
under the year 1052. In 0. Welsh Rhys is writ-
ten Ris and Res, but that the pronunciation of
the initial is correctly given in the 0. Englisli
spelling cannot for a moment be doubted ; for
0. English hi and hr initial had probably the
same sound as in Mod. Icelandic, and I fail to
detect any difference between Icelandic hr and
our rh: my Icelandic friends can pronounce the
consonants in my name just as natives of North
Wales do.

U represented, besides the vowel u, also the
semi -vowel which we write and sound like
English w, as in gnyn, ' white,' and wyneh,
'face.' In a few instances it represents v, as
we have already noticed in connection with the
letter m. .

Before leaving the consonants it should be
mentioned that in the Capella Glosses not only
m, n, r, s are frequently doubled, but also the
mutes c, t, p, especially when they happen to be
final. Ifepp and hepp, now neb, ' any, anybody,'
and heb or eb, ' quoth,' were alluded to in a former
lecture, and to them I should have added Cor-
mac's brace, as proving, beyond doubt, that brdc
was the pronunciation in 0. Welsh of the word
which we now write bragi, ' malt,' and pronounce



(delwedd B6247)


In speaking of the vowels as they appear in
writing, you will have to bear in mind that their
sounds have undergone modifications, in point of
quantity, depending on the nature of the conso-
nants immediately following them. "With this
reserve you may, on the whole, regard 0. Welsh
a, e, 2, 0, ii as pronounced like our modern a, e, i,
0, w. Among the points which require to he
dealt with a little more in detail are the follow-
ing : — (1.) 0. Welsh ? would seerfi to have had,
as far as concerns quality, the same sound as our
y in hyr, ' short,' and dyn, ' man.' This sound
of i may, for the sake of distinction, be called
broad i, and it would appear to have been hardly
such as could be easily distinguished from that of
e and i already noticed as sometimes indiscrimi-
nately written in inscriptions of the Brit-