The Sindarin/Noldorin mutation system is commonly assumed to have arisen
in full parallel with the sound changes that had taken place before the
Sindarin/Noldorin proper system established itself. Now this assumption
rests on the fact that the internal structure of a complex consisting of
a noun and a clitic is quite parallel to that of a standalone word.
While this is, to some extent, true, I wil try to show that the word +
clitic complex is a phonological phrase rather than a phonological word,
and thus their properties differ in certain respects. This is a rather
trivial observation, but I still think it must be taken into
consideration when analysing Sindarin/Noldorin historical phonology.
Bertrand Bellet has dedicated a number of interesting articles to the
history of Sindarin/Noldorin consonant mutations, see, for instance,
Elfling message 12501 or Lambengolmor message 288. Both of these
insightful articles deal, whether primarily or not, with the fact that
the mutation patterns seem to be at variance with the word-internal
Let me consider one example. In the Lambengolmor message referred to
puts forward a number of reconstructed proto-forms for Sindarin nouns in
the genitive qualified by the article. For instance, he cites:
> - "of the word": OS *_ina pettha_ > *_en@ betth@_ > *_en beT_ >
> S *_e-beth_
I find such a derivation plausible (the prehistorical genitive form of
the article *_ina_ (*_inâ_?) is in fact present in the Gnomish Lexicon
(PE11:7)). There is a trouble, as Bertrand himself notes, in the loss of
_n_, which shouldn't have happened, regardless of whether the consonant
were internal or final. He cites, however, the loss of the nasal in an
unstressed prepausal environment in English ('a(n)' vs. 'one'). To this
one can add that nasals tend to be lenes in most languages (Irish and
Scottish Gaelic notwithstanding), and often show a propensity to
disappear (cf. the Welsh _ab_ 'son (in names)' vs. _mab_ 'son' <
*makwos, or the numerous evidence of a weak pronunciation of final
nasals from Latin grammarians). The more interesting question, to my
mind, is the behaviour of that final vowel, which in Bertrand's
reconstruction collapses into a schwa and then disappears.
There can be two different explanations for this. One is regular
word-internal syncope, the other would mean that this is the regular
word-final loss of vowels, and the choice hinges on the presence of a
phonological boundary akin to the one triggering apocope.
The first explanation looks somewhat doubtful to me. Well, we don't know
a lot about the pre-history of Sindarin stress, but I still think the
Sindarin syncope patterns could hardly produce a number of these lost
vowels. First, Sindarin, or rather Noldorin (since most words with
Tolkienian proto-forms come from the Etymologies), syncope tends to
affect first vowels in two adjacent syllables (as in Noldorin *_tarasta_
> _trasta_ (V:391 s.v. TARAS)), or composition vowels in compounds. Thus
I wouldn't expect too many cases of this vowel disappearing to trigger
Anyway, I put rather more faith into the "word-final" explanation. It
gives us however, an apparent contradiction - on the one hand, the final
vowel of the proclitic is subject to "word-final" processes, on the
other hand, the mutation is triggered in seeming accordance with
Now what I am about to suggest is that mutation in Sindarin operates on
the phrasal level, while vowel dropping is a word property.
If this is true, then we should parse the above example as
[PhP [PhW ina ] pettha]
(PhP - phonological phrase, PhW - phonological word; I'm not parsing
_pettha_ as a PhW so as not to clutter up and to show the hierarchical
This allows to account for the fact that _ina_ shows both
"word-internal" and "word-in-its-own-right" properties, since if we
suggest that apocope happens on all right boundaries of *words*, we can
explain the "word-final" properties.
(An alternative account will consider _ina_ a clitic, and the _ina
pettha_ complex the phonological word. The body of the argument is not
affected by this terminological change, since it will still suppose a
special status of _ina_ in the phonology)
Now if we assume that some processes do consider the presence or absence
of a phonological word boundary as a relevant factor, this can have some
important implications for a theory of Sindarin mutations (and for
"practice", of course).
Writes Helge Fauskanger in his article on Sindarin
'The soft mutation corresponds to how certain consonants or consonant
groups develop between vowels.' Building on this, he cites some examples
of medial ND-combinations developing into NN > N, and concludes that the
lenition of original nasalized stops produces simple nasals, since the
processes are exactly parallel.
Faced with the evidence in VT44:21, Helge adds a further comment on
'This system does seem to fit the general phonology best. I would
therefore write _i mas_, not _i mbas_, for "the bread" - irrespective of
Tolkien's curious indecision in this matter.'
Leaving aside the last statement of Helge's, which is not the topic of
this post, I would like to concentrate on his first claim, i. e. that ND
> N 'suits general phonology best'. First, this statement somehow
presupposes a 'better' way for the development of such combinations.
ND-combinations, however, do not always give NN, witness, for example,
Hebrew. Thus the statement is at best suspicious.
I suggest that the lack of ND > NN simplification in "word-initial"
(mutated) position is a special property of, for instance, left
boundaries of PhW's. That ND-groups are permitted in this position is a
fact that Helge himself acknowledges (he lets _e-mbâr nîn_ be an example
of 'mixed mutation'), and so the claim about 'general phonology' appears
to have a rather weak foundation.
Thus, PhW boundary is a factor which influences the non-execution of the
ND > NN simplification. This, by the way, is a good reason for
suggestion underlying ND-groups for *all* Sindarin single nasals
resulting from these groups - we will then need a rule for a
simplifcation if a PhW boundary is absent, or a blocking rule on PhW
boundary). This can lead to a suggestion that mutation and ND > N
simplification are different processes at all, going on different
Pursuing the idea further, we can suggest that left boundaries do appear
to be endowed with special properties with regard to mutations. While a
left boundary of a PhP (not PhW - the PhW left boundary is in fact the
only natural context for mutation!) certainly does not preclude mutation
(witness Direct Object lenition or things like _gûren bêd_[*]), it does
lay severe restrictions thereon, requiring a syntactic trigger rather
than a purely phonological one. This, once again, can be considered as
evidence of the fact that what is termed a single process ('mutation')
can in fact be a sum of a number of processes operating on different
[*] On a rather unrelated note, one wonders if the two words do not form
a single PhP by virtue of _gûren_ being fronted for emphasis, and thus
the logical stress relocating to that word. The assumption of fronting
does require the assumption of unmarked VSO order, which the present
author rather favours. It is, however, not directly relevant to the
question at hand)
- Sindarin does possess a distinction between phonological words and
phonological phrases (but see above on terminology), and possibly other
gradations not considered here.
- Different phonological processes take into account different levels of
- Mutations can be decomposed into a number of processes operating on
- There is nothing inherently wrong with a form like _i mbas_ for 'the
bread' (which did not require the above rant, since it was written by
Tolkien himself! :-)).
These are rather unpolished thoughts, though.
Pavel Iosad pavel_iosad@...
Nid byd, byd heb wybodaeth