In Noldorin and Sindarin, etymological unvoiced stops generally appear to
have been vocalised in implosive positions before other stops or s, forming
diphthongs with the preceding vowel -- probably via an intermediary spirantal
stage, as general phonetics and the parallel of primary world languages suggest.

Quenya cognates generally preserve a consonant cluster, though sometimes
modified. The process is on the whole better attested in Noldorin than in
Sindarin. Examples (¤ = Tolkien's reconstruction, * = other reconstruction) :

¤_ektele_ "spring, issue of water, well" > N / S _eithel_, Q _ehtele_
(V:363, S:358, 360, TC:187)
¤_okta_ "war" > N _auth, Q _ohta_ (V:365, 379)
*_leptā-_ "pick with the fingers" > S _leutha-_, Q _lepta-_ (VT47:10) (the
diphthong eu is problematic)
*_taksź_ "nail" > N _taes_, Q _takse_ (V:389, 390)
*_apsā_ "cooked food, meat" > N _aes_, Q _apsa_ (V:349)

It also applies to g before liquids and nasals -- probably it was first
lenited to 3 (=spirant g) and then vocalised.

¤_magla_ "stain" > N _mael_ (V:386)
¤_magrā_ "useful, fit, good"> N _maer_, Q _mįra_ (V:371)
¤_magnā_ "skilled" > N _maen_ (V:371)
¤_sagmā_ "poison" > N _saew_, Q _sangwa_ (V:385)

This is a fairly common change in several primary world languages :

i. English is a case with the evolution of the former velar fricative spelt
gh in _eight_, _daughter_

ii. It is also found in (Western) Romance, e.g. Latin _lactem_, _laxare_ >
French _lait_ "milk", _laisser_ "leave, let"

iii. More interestingly for Sindarin, it is well attested in Brythonic
language, among them Welsh. The rules are the following (PIE =
Proto-Indo-European, CC = Common Celtic) :

PIE kt, pt > CC cht preserved in Gaelic, vocalised in Brythonic languages.
E. g. *kaptos > Irish _cacht_ "servant", Welsh _caeth_ "slave, captive" --
compare Latin _captus_ "taken"

PIE ks, ps > CC chs preserved in Gaulish, > s in Gaelic, > ch in Brythonic.
E. g. *oupselos > Gaulish _uxello-_ (Latinised spelling), Irish _uasal_, W
_uchel_, all meaning "high" -- compare Greek _hypsźlos_

(Adapted from : Henry Lewis & Holger Pedersen, _A Concise Comparative Celtic
Grammar_, 3rd edition, Vandenhoek & Ruprecht, Göttingen, 1974)

Latin ks (written x of course) however became a secondary xs which developed
differently, with x vocalised and s remaining : Latin _crux_ > Welsh _croes_
"cross", Latin _Saxō_ > Welsh _Sais_ "English"

This has been known for long. What I would like to point out in this message
is as series of regular differences between Noldorin and Sindarin in the
development of these clusters, which imply that Tolkien had partly revised his
ideas on the historical phonology of these two stages of his Welsh-sounding
Elvish language.
_____________________________________________________________________

1) okt, ukt > auth, uth in N but oeth, uith in S
N and S agree that vocalisation after _e_ gives _ei_ (>_ai_ in final
syllables in S) and after _a_ gives _ae_.

¤_ektele_ "spring, issue of water, well" > N / S _eithel_, Q _ehtele_
(V:363, S:358, 360, TC:187)
¤_wegtź_ > N _gweith_ "manhood, manpower, troop of able-bodied men, host,
regiment" / S _gwaith_ "people" (S:359, V:398, VT46:21)
¤_magrā_ "useful, fit, good"> N _maer_, Q _mįra_ (V:371)
S _śthaes_ "inducement to do wrong" dissimilated from _śthaeth_ compare Q
_śsahtie_ (VT44:30)

After _o_ the vocalisation gives _au_ in N

*_loksź_ "hair" > N _lhaws_, Q _lokse_ (V:370)
¤_okta_ "war" > N _auth_, Q _ohta_ (V:365, 379)

But the vocalisation gives oe in the only Sindarin example I can remember
of

¤_logna_ "soaking wet, swamped" > S _loen_ (VT42:10)

After _u_, the vocalisation is generally _ū_ or _u_ in Noldorin.

¤_lugni_ "blue" > N lhūn, Q lśne (V:370)
*_luktā-_ "enchant" > N lhūtha-, Q luhta- (V:370 corrected in VT45:29)
¤_suglu_ "goblet" > N sūl, Q sślo (V:388)
*_suktu_ or *_suktō_ "draught" > N sūth, Q suhto (V:388)
ON _tulugme_ "support, prop" > N tulu (V:395)

Sometimes it is in _au,_ probably because here _u_ was opened to _o_ by
metaphony before the stop was vocalised.

*_suktā-_ "drain" > N _sautha-_ (V:388)
¤_tupsź_ "thatch" > N _taus_, Q _tupse_ (V:395)

(Why this happened in *_suktā-_ and not in *_luktā-_ is not clear :
different sequences of changes ? inconsistency from Tolkien ?)

However _ui_ appears at least in the N _iuitho_ "?to enjoy [possibly a
misreading for "to employ"]" (V:400, VT46:23) from the root YUK; the same
evolution is seen in the related word _iuith_ "use". Perhaps the initial i
influenced
the vocalisation; alternatively Tolkien might have begun to introduce the
later Sindarin system. For in Sindarin, ui is seen:

*_gruktā-_ "terrify" > S gruitha- (XI:415)
*_luktiā-_ "quench" > S_ #luithia-_ in _uluithiad_ "unquenchable", literally
"without quenching" (IX:62)
*_nuktā-_ "stunt" > S _nuitha-_, Q _nuhta-_ (XI:413)

It is also significant that Tolkien changed the etymology of _Lśthien_ from
¤_luktiēnē_ "enchantress" (V:370, VT45:29) to a derivative of the root
LOTH
meaning "daughter of flowers" (PE17:161). Was it because the old etymology
would now have given *_Luithien_ ?

____________________________________________________________________

2) ps, ks > diphthong + s in N but f, ch in S

In Noldorin, we see that the stop is vocalised in the clusters ps, ks:

*_apsā_ "cooked food, meat" > N _aes_, Q _apsa_ (V:349)
*_khelkaraksź_ > N _Helcharaes_, Q _Helkarakse_ (V:362)
*_lapsź "babe" > N _lhaes_, Q _lapse_ (V:367)
*_loksź_ "hair" > N _lhaws_, Q _lokse_ (V:370)
*_taksź_ "nail" > N _taes_, Q _takse_ (V:389, 390)
¤_tupsź_ "thatch" > N _taus_, Q _tupse_ (V:395)

But PE17:134 says about Northern Sindarin :

...tt, pp, kk > t, p, k medially. ch, ž, f only derived from kh, th, ph and
ks, ts, ps...

As I understand it, it implies that, while the (well-known) tt, pp, kk > ž,
f, ch were specifically Southern Sindarin (the kind we are most familiar of),
the changes kh, th, ph & ks, ts, ps > ch, ž, f were GENERAL, found in all
dialects of Sindarin.

We can note that in Noldorin already ts > tth > ž :

ON _litse_ > ON _litthe_ > N _lith_ "sand" (V :369), in S _lith_ "ash,
dust" (S:361), compare Q _litse_
_etsiri_ > N _ethir_ "mouth of a river, estuary" (V:356), also in S (LR/II
ch. 10, S:364), compare Q _etsir_

We can probably generalize this pattern "unvoiced stop+s > double aspirated
stop > unvoiced fricative" in Sindarin :

ps > pph > f
ks > kkh > ch

The idea was already voiced by David Salo in _A Gateway to Sindarin_ p.
71 : starting from the correspondence of S _carach_ "jaw" (S:429, RC:607) / N
_#caraes_ "jagged hedge of spikes" (V:362) -- to be compared with Q _#caraxe_
from the famous name _Helkarakse_ -- he wrote :

...In a Sindarin dialect of unknown affiliation the sequences ks and ps
became kkh and pph respectively and eventually x [=ch], f. (...) It is possible
that this is the normal Sindarin development and that the forms in which ks
and ps changed to *is > es are actually Noldorin...

He appears to have been right. We can add as new evidence, beside Tolkien's
word, the correspondence Q _axo_ / S _achad_ "neck" (RC:537, PE17:146) from a
root AKAS.

It is noteworthy that Tolkien successively imitated the developments of the
two kinds of ks, inherited and borrowed from Latin, that are found in Welsh.

Yours,
Bertrand Bellet