The origin of mixed mutation
- Greetings !
This post is an attempt to track back the origins of mixed mutation in
Sindarin, partly with what we can deduce from the internal evolution of
the language, partly thanks to the evidence of Welsh and Breton.
I will used the following notations: N == velar nasal (Tolkien often used
an n with tilde); T == unvoiced dental spirant (Sindarin th); D == voiced
dental spirant (Sindarin dh) ; 3 == voiced velar spirant (spirant g). I
have preferred accents and circumflexes to doubling and tripling of
vowels, following the advice given in the list's guidelines.
While the grammatical uses of mutations in Sindarin are not always easy
to determine, their ultimate phonetic origin is usually quite transparent:
they are the reflects in external sandhi of sound changes regularly noted
inside unitary words: so for soft mutation (arising from internal
slackening of s > h; internal voicing of unvoiced stops and spirantisation
of voiced stops and m, or lenition proper; initial strengthening of w- >
gw-; and initial weakening of ch- > h-), the so-called stop mutation (from
the aspiration of unvoiced stops when geminate or after a consonant) and
the rather hypothetical liquid mutation (aspiration of unvoiced stops and
spirantisation of voiced stops and m after liquids). Nasal mutation is a
bit more complicated - difficulties arise with unvoiced stops - but
reflects more or less clearly the evolution of clusters including a nasal.
However, mixed mutation appears puzzling at first sight. Let us summarise
what we know.
- Unvoiced stops appear to be voiced : _Narn e-Dinúviel_ "Tale of the
Nightingale" (MR:373), from Tinúviel, the nickname of Lúthien ; _erin
dolothen Ethuil_ "on the eighth day of Spring" (King's Letter, SD:128-
129), cf. _toloth_ "eight" in LR:394.
- Voiced stops from primitive *b-, *d-, *g- are generally unchanged:
_ben genediad Drannail_ "in the Shire-reckoning" in the King's Letter,
cf. _gonod-_ "count, reckon" (LR:378,379); _Narn e-Dant Gondolin_ "Tale
of the Fall of Gondolin" (MR:373), the basic form being undoubtly _dant_
"fall", attested in the season name Dantilais "leaf-fall == autumn",
PM:135, coming from the root DAT (LR:354), to compare with the Quenya
_lante_ in Maglor's lament _Noldolante_ "Fall of the Noldor" (Silm.
- gw- (from primitive *w-) is unchanged: _erin Gwirith edwen_ "[on] the
second day of April" in the King's Letter; the name of the month is
given as _Gwirith_ in LotR, Appendix D.
- Voiced stops from primitive *mb-, *nd-, *Ng- "revert" to their
primitive prenasalised form: _ ered e_mbar nín_ "[the] mountains of my
home" (UT:40,54) and _Narn e-mbar Hador_ "Tale of the house of Hador"
(MR:373) with _bar_ "house, home" from _mbár_ (read _mbar_? Silm.
Appendix); _Bar-en-Danwedh_ "House of Ransom" (Silm. ch. 21), a possible
spelling variant of _Bar-e-Ndanwedh_, the last word probably including a
product of the root NDAN "back", cf. for instance PM:395 ; _Haudh-en-
Ndengin_ "Hill of the Slain" (Silm. ch. 20) with a descendant of the root
NDAK "slay" (Silm. Appendix entry "dagor"). We also see this treatment in
_Taur-e-Ndaedelos_ "Forest of the Great Fear" (LotR Appendix F), though
_daedelos / daedhelos_ rather comes from a simple D-stem (LR:354, 355).
- Fewer examples are available for other cases: at least f and n are
unaffected: _Taur-en-Faroth_ (Silm. Appendix entry "faroth"), _Haudh-en-
Nirnaeth_ "Mound of Tears" (Silm. ch. 20); _nirnaeth_ is attested
initially in the famous battle name _Nirnaeth Arnoediad_ (Silm. ch. 20).
So we have something like lenition for unvoiced stops, nothing for
"normal" voiced stops and something like nasal mutation for voiced stops
from old prenasalised voiced stops. How can we explain these confusing
phonetic developments? I suggest we consider a process called provection.
Kenneth Jackson in _Language and History in Early Britain_ (Edinburgh
University Press, 1953), pp. 561-62, describes it so for Welsh:
When two consonants came together secondarily in WCB [Welsh-
Cornish-Breton], either by syncope of the intervening vowel or
by the creation of a new compound, or even at the end of one word
and the beginning of the next, there may take place the
accommodation of articulation such as the unvoicing of voiced
sounds or the change of spirants to stops, which is called
provection. The result may be to nullify the effects of lenition:
as, for instance, where the Latin _benedictio_, Late Brit.
_*beneDixtî_, becoming _*ben'Dixt_ by loss of final syllables and
syncope, appears in W. _bendith_ with its lenited _D_ once more a
[D is the dental voiced spirant, a d with a bar in the original; x is
the Ach-Laut, a chi in the original; the circumflex is a macron is the
Jackson also writes: "CB. have a much more thorough-going system, built up
into what amounts to a regular morphophonological external sandhi".
Breton has indeed two kinds of provective mutations:
- one called in French "mutation durcissante", i.e. hardening mutation,
or provection; it unvoices voiced stops;
- another called "mutation mixte", i.e. mixed mutation, or leniprovection;
it is sometimes similar to lenition, sometimes to provection. Kenneth
Jackson analyses it in "A Historical Phonology of Breton" (The Dublin
Institute for Advanced Studies, 1967). Very grossly, it comes from a
complex succession of lenition, spirantisation, provection and new
lenition (a secondary voicing of spirants occurring in many Breton
dialects), with much dialectal variation.
The latter may indeed give ideas for Sindarin. Now the question naturally
arises to know whether provection is attested in Sindarin. I think it
may be; it may explain sometimes why lenition fails to occur where it
might be expected, though it is necessarily somewhat speculative since
lenition rules are not completely understood. In particular, it is
noteworthy that lenition fails to occur in second element of compounds
after a nasal, though these are often lenited. Here are a few examples
(the list is not exhaustive):
- _Baranduin_ (LotR map and Appendix F): _baran_ "brown" + _duin_ "(long
and large) river"
- _Annúminas_ (LotR map): _Annûn_ "west, sunset" + _minas_ "tower"; not
- _Glanduin_ (UT:264, LotR map): _glan-_ "border" + _duin_ "(long and
- _Thuringwethil_ (Silm. Appendix): _thurin_ "secret, hidden" + _gwethil_
"she-of-shadow" (certainly from _gwath_ "shadow")
- _Glamdring_ (the Hobbit, LotR): _glam_ "uproar > orcs > foes" + _dring_
- _dangweth_ "answer" from _*ndanagwetha_ (PM:395).
Perhaps also some failed lenitions like in _Lond Daer_ "Great Harbour",
but this is doubtful: we have also _Eryn Vorn_ "Dark Wood" (LotR map).
In Didier Willis' Sindarin dictionary, David Salo puts forward the
following etymology for the genitival article _en_: it would come from an
earlier _*ina_, i.e. the deictic particle (or already article?) _i_ +_na_
used as a genitive sign, as is well-known in Noldorin (cf. names like
_Taur-na-Fuin_, _Dor-na-Fauglith_, and the note in the Etymologies, root
NA, LR:374) and, less so, in Sindarin (cf. names like _Taur-na-Neldor_ and
_Orod-na-Thôn_ sung by Treebeard in LotR book III ch. 4). So _Haudh-en-
Ndengin_ would be etymologically "mound the (one) of slains". This
etymology goes quite well with the idea of a mixed mutation caused by
_lenition followed by provection_. Let us imagine a few home-made
examples (@ == schwa; 3 == back voiced spirant):
- "of the tree": Old Sindarin *_ina galada_ > [lenition and umlauts]
*_en@ 3alaD@_ > [loss of final vowels] *_en 3alaD_ > [provection] *_en
galaD_ > Sindarin *_e-galadh_
- "of the word": OS *_ina pettha_ > *_en@ betth@_ > *_en beT_ >
- "of the land": OS *_ina ndore_ > *_en@ ndor@ > *_en ndor_ >
I must admit that there is an important weakness in this: why would _en_
be reduced to _e_? I cannot explain it satisfactorily. Nevertheless,
special reductions of this kind are not unknown in unstressed grammatical
words. The English indefinite article "a(n)" - with the same etymology
as "one" - provides an interesting parallel. Also, the plural article
_in_, used as a plural genitive article too, regularly appears under a
reduced form _i_, especially before stops, because the -n is swallowed in
nasal mutation. Perhaps it influenced somewhat the genitive singular
article _en_, whose reduced form _e_ would have originally arisen in cases
like *_en ndor_ > S. *_e-ndôr_? There might in fact have been some
variations (externally or internally) if we consider the above quoted form
_Taur-e-Ndaedelos_ as an alternative spelling for _Taur-en-Daedelos_.
As for the suffixed article _-in_, mixed mutation could be explained in
the same way if the primitive form was _*-inV_ (V == a vowel, not == a
because it would have caused a-umlaut).
The form _Taur-in-Duinath_ "Forest between Rivers" (Silm. map) might also
be explained thus. _Im_ "between" is presumably related with the Quenya
word _imbe_ of the same meaning appearing in "Namárië". The Old Sindarin
form must have been similar, and since its ended in a vowel it is
surprising that its descendant does not trigger the lenition of _Duinath_.
But the final -m may have had a provective effect, turning back voiced
spirants to stops. Hence _im_ would trigger mixed mutation. However, the
evidence available today does not allow us to check this, and other
theories are equally possible.
Lastly, the fact that the mixed mutation of gw- appears to be gw- may be
interesting in establishing the chronology of sound changes producing
Sindarin: if it was not influenced by the mutation of g- (that's a big
"if"), it points out that the strengthening of initial w- > gw- is
anterior to lenition, the evolution being OS *ina w- > *ina gw- >
*en@ 3w- > *en 3w- > *en gw- > S e gw-. If it had been posterior, we
would have conceivably OS *ina w- > *en@ w- > *en w-, which would give
**en w- or perhaps **e w- in Sindarin. But see the "if" above... we
cannot be sure.
Thank you for your attention. Comments are welcome.
[Re: _Annúminas_, this may be an example of the orthographical
representation of long _mm_ (< *_n-m_) as simple _m_; cf. Tolkien's
comments on the representation in Appendix E: Appendix E: "_mb_ became
_m_ in all cases, but still counted as a long consonant for purposes of
stress ..., and is thus written _mm_ in cases where otherwise the stress
might be in doubt". CFH]
[Non-text portions of this message have been removed]