|Bertrand Bellet — August 2005|
|Theoretical Articles: A comprehensive knowledge of the works of J.R.R. Tolkien is needed to fully understand articles in this category, the subjects treated being studied in minute detail by their authors.|
owel affection is a prominent phonological process in the morphology of J. R. R. Tolkien’s Sindarin and Noldorin. This feature appeared in the late stages of his Gnomish and remained in his series of Welsh-inspired languages ever after. Its various kinds and the terminology used to describe them are first introduced cross-linguistically, with special reference to Celtic and Germanic languages. Patterns of vowel alternations in Sindarin and Noldorin from the Etymologies onwards are then examined synchronically: plural formation, variation between related morphemes, alternations linked to affixation and composition are successively considered. From these data a model is elaborated to explain the diachronic development of vowel affection, by means of comparison with Quenya and internal reconstruction, as well as examination of Tolkien’s own indications. The model includes five stages : lowering of high vowels by A-metaphony, conditioned lowering of U to O, raising of mid vowels by I-metaphony, fronting of back vowels by a later I-metaphony, and finally breaking caused by palatalised consonants. In each case a brief phonetic explanation is given with a list of relevant examples; difficult cases and limits of the model are also discussed. Finally, this reconstructed phonological history is compared to the development of Welsh and the Germanic languages, in order to appreciate how far Tolkien used them as models when he built his languages.
ne of the most striking features of Sindarin, J. R. R. Tolkien’s famous Welsh-inspired language, is the prominence of vowel alternations in its morphology, both inflexional and derivational. Vowel changes provide the most common and productive way of forming the plural of nouns and adjectives, and are regularly triggered by the addition of certain suffixes to radicals. This feature is deeply rooted in the external history of Tolkien’s Welsh-inspired languages. It does not yet appear in the oldest layer of Gnomish as seen in the Gnomish Lexicon of 1917, but vowel alternations are used to mark the plural of nouns in later additional slips published in Parma Eldalamberon #13. Vowel affection is well developed in the Early Noldorin Grammar that follows (ibid.). Tolkien explained in this grammar how these alternations arose historically by vowel affection, but the scenario as presented in this early document cannot apply unchanged to the later Noldorin and to Sindarin. Firstly, Tolkien slightly modified the patterns over the years; secondly, and more importantly, he totally changed his views about the primitive Elvish language while he wrote the Etymologies, our greatest source of Elvish vocabulary. This means that his ideas about the development of his languages must have been greatly modified at that time. Therefore we need a study of vowel affection for the later external stages of Tolkien’s Welsh-inspired languages, for no one of his hand has been published.
This is the purpose of this essay to provide such a survey for Sindarin and the externally later stages of Noldorin: briefly, for all stages of this branch of Elvish contemporaneous or later than the Etymologies (roughly from the mid-thirties onwards). Vowel affection will first be presented in general terms and exemplified in a variety of languages; special attention will be paid to Welsh and Germanic languages, since Tolkien knew them very well and evidently took them as models (this is well-known for Welsh, less so for Germanic languages1)). Then, patterns of vowel alternation will be examined in synchrony, and the various phonetic processes that can be distinguished in them will be set out. Finally, a diachronic model for the development of vowel affection will be proposed and compared back with Welsh and Germanic languages. Several methods are available to achieve this purpose. The first one is of course a close examination of Tolkien’s notes. As a philologist, diachrony was very important to him, and he was visibly deeply involved in developing a plausible and coherent history to his languages. As a consequence, he left many statements about this topic, and also provided us a number of reconstructions of early stages of Elvish (Primitive Quendian, the later Common Eldarin, the still later Old Noldorin which is the nearest ancestor of Noldorin), which are often very enlightening.
The second kind of methods that can be used are the two great tools of historical linguistics: comparative and internal reconstruction. Comparative reconstruction uses regular correspondences of sounds between genetically related languages to deduce forms of their common ancestor. Each correspondence allows determining the evolution of a phoneme of the parent language in one given phonetic environment: thus we can retrieve the allophones of the parent language. Taking the various phonetic environments into account, provided that we have enough correspondences, we can deduce the primitive phonemes and the sound laws operating between the parent and the descendant languages. This method does not indicate the phonetic value of the primitive phonemes, but guesses can be made on the basis of the phonemes of the descendant languages and considering what changes are the most probable. A cross-linguistic regard – linguistic typology – is very helpful there. In our case, comparison will mainly be made with Quenya, sometimes with the Telerin of Aman too; this allows us to go back to Common Eldarin. Internal reconstruction studies allomorphy (conditioned morpheme variation) in one same language to retrieve earlier forms of this language; it is based on the assumption that allomorphs split from an ancestral single morpheme by sound changes operating in different phonetic contexts. It will be used quite often to look at specifically Sindarin (or Noldorin) phenomena that cannot be retrieved by comparative reconstruction. Internal reconstruction is nevertheless limited: it can retrieve neither “see-saw” changes (A > B > A) nor complete mergers (A > C and B > C), since it works on differences surviving in one stage of a language. Aditionally, used in a extreme way, internal reconstruction assumes a primitive language without morphophonemics, which can be questioned.
Finally, languages Sindarin is known to have been built upon are useful as a heuristic tool: they can be scrutinised to suggest patterns of change. Of course, they must be confirmed afterwards by pure examination of the Sindarin/Noldorin material.
All quotations in Tolkien’s languages will be referred to their sources according to the
system used on the Tolkiendil website2). I will also use the following abbreviations.
¤ Tolkien’s recontruction
* other reconstruction
** wrong form
# deduced form of a word only attested inflected or compounded
† archaic or poetic
< derives from
> evolves into
PE Primitive Elvish
ON Old Noldorin
EN Exilic Noldorin
T Telerin (of Aman)
When reconstructing archaic stages of Sindarin and Noldorin I will use the following symbols in addition to those commonly found in Tolkien’s writings: þ and ð for voiceless and voiced dental spirants, i.e. Sindarin’s th and dh; χ for the voiceless velar spirant, i.e. Sindarin’s ch, ȝ for a voiced velar spirant lost in later Sindarin/Noldorin, that Tolkien sometimes spelt gh when he needed to; and µ for the nasal spirant resulting from the lenition of m, for which Tolkien used the digraph mh. As Tolkien often did, the ligature œ will be used to denote a mid-open front rounded vowel, like French eu or German ö, resulting from the fronting of o. A raised i will denote a reduced, non-syllabic i. Additionally, to prevent confusion between y used as a (high front rounded) vowel and a (palatal) glide, I will reserve y for the vowel, use j for the glide and alter Tolkien’s spellings accordingly when needed.
t is a common linguistic development that vowels can be affected by their neighbours. In general terms, such a distance assimilation of a phoneme to another not contiguous one is called dilatation; it is not restricted to vowels. Concerning vowels these assimilative processes have received a large array of names, which can make the terminology of this area of phonology rather confusing. So it is perhaps not superfluous in this article to state precisely which terms will be used for which phenomena. In Sindarin, we see that a vowel can be affected by another one following it in the word. This also occurs prominently, among other languages, in the Celtic and Germanic families, which were well known to Tolkien. In German, this is called Umlaut - the word refers simultaneously to the diachronic change, to the vowel alternations that have arisen therefrom and become part of the morphophonology of the modern language, and to the diacritic used to denote it. The word had been calqued with Greek elements to form the term metaphony, that I will use in this article. More limited terms are in use to denote special types of metaphony. They often refer to the move observed in the point of articulation of the vowel. We thus encounter lowering, raising, fronting. The affecting element can also be specified, and we get I-affection, I-metaphony, and so on.
Metaphony by A tends to lower high vowels like i and u into the direction of e and o. It can be found for instance in the West and North Germanic languages:
Its workings are especially clear in Old High German, whose morphology displays regular i / e and u / o alternations; see for instance ih hilfu “I help” vs. helfan “to help”, sia hulfun “they helped” vs. giholfan “helped”. Traces can still be found in Modern German, for instance sie wurden “they became” vs. geworden “become”, Hilfe “help” vs. helfen “to help”.
In the development of Welsh, final -ā lowered a short i and u to e and o respectively in the preceding syllable. In the modern language, it notably results in a gender distinction carried by the stem vowel in some adjectives. The masculine forms have y or w, while the feminine forms have e or o. This goes back to an opposition in the parent language between a masculine ending in *-os (not causing metaphony) and a feminine in *-ā (causing metaphony):
Metaphony by I tends to raise and front other vowels. It had dramatic consequences in the morphology of German, English and Scandinavian languages, who all display evident traces of I-metaphony. Some examples with Old English:
Plural of some nouns (originally with an *-iz ending, later lost):
Comparatives and superlatives (originally with an *-ira and *-ista endings):
Conversion of adjectives into nouns with the suffix *-iþō (originally):
I-metaphony is also of considerable importance in the Brythonic languages; here are some morphological alternations produced by it in Welsh.
Plural of nouns (originally with an -ī ending, later lost):
3rd person sg. present indicative (in older and literary language):
Metaphony caused by various suffixes containing y or i:
Metaphony by U tends to raise, back and round other vowels. It is widely attested with many variations among North Germanic languages. Some examples:
Breaking is a kind of dilatation turning a vowel into a diphthong under the influence of a neighbouring sound, which can be a vowel or a consonant:
Old English: in the West-Saxon dialect especially, front vowels are turned into diphthongs by a following h, r + consonant, and l + consonant in some cases:
Old English: short front vowels are turned into short diphthongs by a back vowel in the next syllable (this is called “back umlaut” or “U-umlaut” in Old English primers, though it is quite different to the Old Norse phenomenon of the same name)
Old Norse : e > ja (whence jö6) by later U-metaphony) when the following syllable contained a back vowel (with restrictions)
Breaking is especially encountered when discussing Germanic languages, but similar phenomena can be found elsewhere.
Rumanian: e and o are broken to ea and oa by a following a or ă; this results in numerous morphological alternations. Some examples:
Welsh: a and e sometimes shift to ei when a suffix with consonantal i is added:
Epenthesis is a very general term referring to the insertion of a non etymological phoneme into a word for a variety of reasons: euphony, ease of articulation, analogy, etc. The term of vowel affection is often encountered in English, especially among Celticists, to name the influence exerted on a vowel by a following one; it is thus more or less a synonym of metaphony. In this article I will use it loosely as a cover term for every kind of conditioned vowel change.
t is well known that most plurals in Sindarin and Noldorin are build by means of vowel changes. They can be traced back to vowel affection triggered by a former plural ending -i, attested in the Old Noldorin of the Etymologies: boron “faithful vassal” pl. boroni (LRW:353), poto “animal’s foot” pl. poti (LRW:382), toron “brother” pl. toroni (LRW:393). The vowel alternations between the singular and the plural of nouns and adjectives that were produced thereby were later grammaticalised as a productive mark for number in Noldorin, and evidently Sindarin. Studies have already been devoted to these patterns8), so I will simply summarise their results in tables. Unattested cases are left in blank.
|Singular||Plural in final syllables||Plural in non-final syllables||Remarks|
|a||ei||in alph “swan” pl. eilph (UT:245)|
|a||e||before r + cons., before ng in ered and ened|
|a||a||probable in annui (SD:129-31)|
|o||y||œ > later e|
|au||oe||Nibin-noeg “Petty-dwarves” (WJ:458) must contain the pl. of naug “dwarf” (WJ:388)|
|ei||î||in feir pl. fîr (WJ:387)|
For Noldorin in the Etymologies:
|Singular||Plural in final syllables||Plural in non-final syllables||Remarks|
|a||ei > e||e||general case in polysyllables|
|a||e||before consonant clusters|
|a||ai||in aran “king” pl. erain (S pattern)|
|e||ei||in cef pl. ceif “earth” (LRW:363)|
|o||œi > ei > e||œ > later e||in polysyllables|
|o||ui > y||œ > later e||in polysyllables|
|o||ui > y||o||in polysyllables when o < *ā|
utside plurals, internal vowel variation is also sporadically seen between evidently related morphemes. We have notably a series of pairs made of a noun with i or u as root vowel and a related adjective e or o respectively. Many have Quenya cognates. The Old Noldorin form is sometimes mentioned too.
The Old Noldorin or Primitive Elvish forms and the Quenya cognates show that the i / u is original, and that in the Sindarin / Noldorin pairs the adjective must have ended in -a, while the noun ended in another vowel. Most Quenya adjectives end in -a, and we can surmise that it was already true in Common Eldarin (in the form *-ā).
Other cases of alternation between e / i and o / u include the following.
There is series of alternations between i or í on the one hand, ei (Noldorin) or ai (Sindarin) on the other hand.
Comparison with Quenya shows that the ei / ai comes from an earlier sequence *-iCjā (where C = a consonant).
owel affection is also seen to be triggered by the addition of several suffixes containing i. There are many examples. Two rare suffixes -i are known in Sindarin : an adjectival suffix (from *-îia) and a collective (from *-iie). The name of the river Serni, made on sarn “ pebble ”, must contain one of the two (VT42:10-1).
In Noldorin, a suffix -i (from ON -ie) is used to build the infinitives of basic verbs. The original vowel appears elsewhere in the conjugation, in related words and in Quenya cognates.
Another Noldorin suffix for infinitives is -io (from ON *-iōbe). The original vowel can be retrieved by comparison with Quenya or internal reconstruction.
A personal suffix -il is found both in Sindarin and Noldorin. It is especially – though not exclusively – used to build feminines. When it causes affection, the original vowel can usually be retrieved by internal reconstruction.
A diminutive and singulative suffix -ig is attested in Noldorin and Sindarin.
An adjectival suffix -in is sparingly attested in Sindarin and Noldorin.
indarin and Noldorin have a quite common ending -dir. It was originally an independant word: N dîr (LRW:352, 354), indirectly attested in S dírnaith (UT:282) “man-spearhead”, the name of a Númenorean military formation. But as an independant word it had become archaic (this is explicitly indicated in the Etymologies); it seems to have survived chiefly in composition. Most often it does not trigger vowel affection: cf. names like Mithrandir, Curunír, Brandir, Handir, Haldir, Borondir etc. But it does in two agental formations apparently made of the gerund of verbs + dîr (with vowel shortened to -dir); in these words it approaches the statute of a suffix.
Adjectives with the suffix -en display the same metaphonies as in their plural form when used as first members of compounds, even when the meaning is not plural.
In some compounds the first element shows metaphony caused by an i in the second element. These compounds were probably no longer felt as such: some contain elements lost as individual words or display phonetic alterations making them opaque. In most compounds however, metaphony is not found, either because they were formed after metaphony became inactive, or because the unaffected form of the first element was restored after the independant form.
In a few compounds where the second element begins in prevocalic i-, metaphony occurs and produces a diphthong.
n the alternations we have seen till now we can extract several types of vowel affection. Alternations between e / i and o / u, that appear from Old Noldorin evidence and comparison with Quenya to be linked to the presence of a former final -a, point towards a stage where a lowered short high vowels. We have thus to posit an A-metaphony somewhere in the development of Sindarin. However, comparison with Quenya shows a number of correspondences between Q u / S o that are not linked to the presence of a final -a:
We have consequently to posit another mechanism able to turn u to o in Sindarin.
The affection by i displays the most complex patterns.
|Vowel||Ultimate affection||Non ultimate affection|
|a||ei / ai / e||e|
|o||y (N also œi > ei)||œ > e|
|ó / au||oe (N ui)||–|
The ultimate I-metaphony of A looks especially puzzling with three possibilities. However, ai and ei are essentially variants of one same phoneme in Sindarin : ai appears in final syllables and ei otherwise ; compare for instance teithant “wrote” and andaith “long mark”, or erain “kings” and Ereinion “Scion of Kings”. In Noldorin this variation does not appear and ei is found throughout – it seems that Tolkien instead played with the idea that ai for ei was a dialectal peculiarity of the Feanorians (see the entry MAD in VT45:30). As for e, its appearance seems to be linked to special phonetic shapes. In Noldorin it can also be a late modification of unstressed ei. Therefore it would seem that there is some kind of complementary distribution between those three variants for the ultimate affection of a. It could be considered, as Helge K. Fauskanger suggests it in his article Sindarin - The Noble Tongue, that the original affection of a is to ei and that ai and e are later modifications. If we observe the various changes that i can trigger, we note that:
A relatively simple way to handle this intricate pattern is to consider that there was not only one wave of I-affection in Sindarin, but that we observe in synchrony the superposition of several successive alterations, each of one affecting the preceding vowels in a different way. In this hypothesis we would need three waves of I-affection: one to explain the fronting, one to explain the enhanced raising of e and o in final syllables, and one to explain the breaking. This makes for a total of five waves of affection:
To order these changes, we can sometimes rely on clues given by Tolkien himself. In WJ:400 notably, he gives us the reconstructed development of the Sindarin form of the name of Ossë (conserved in the S names Yssion and Gaerys): ¤ossī > ¤ussi > ¤yssi. We see here that I-metaphony raised o > u before fronting it to y. Otherwise we can deduce it logically by internal reconstruction applied to words that underwent several of these changes. We can test several orders and see which one leads to the attested form in the most plausible way. For instance we have in VT46:7 the ON nuhina > EN nohen, noen “wise, sensible”, with both A-metaphony (i > e) and lowering of u to o by e. With the order “U lowering, then A-metaphony” we would have nuhina > *nohina > *nohena > nohen; with “A-metaphony” first we have nuhina > *nuhena > *nohena > nohen. The first is phonetically more difficult because we have to posit that u is lowered to o before a following i, while we know (see just above) that i actually had the opposite effect. The second ordering is thus to be preferred.
A final -a lowers i and u to e and o in the penultimate syllable.
The final sequence -ia has the same effects9).
It can be asked if A-metaphony affects vowels other than in the penultimate syllable. It is not the case for I, generally.
Cf. also the reversion of -en adjectives to -in in composition, that has already been observed (Celebrimbor, Melthinorn etc…). It testifies for a pattern in which internal a did not cause metaphony, by contrast with final a. This survived as an “anomaly” in compounds incorporating an -en adjective. Whether this alternation was still productive in Noldorin / Sindarin cannot be ascertained however: it is quite possible that these names were archaic survivals.
There are however a few words where i is changed to e in the penultimate syllable: estent “?very short” (UT:188, WJ:311-5) and gwelwen “lower air = Q vilwa” (LRW:398). The first is interpreted by Patrick Wynne and Carl Hostetter10)) as a variant of thent “short” (LRW:388), and “can be explained as derived from the same form by the addition of prefixed sundóma or base-vowel (which prefix, like the sundóma in both forms thent and estent, has been lowered by A-affection from *i to e-): i.e. *stintā > *istintā > estent”. There is a difficulty, since the examples above show that A-metaphony affected penultimate syllables only. But since thent and estent must ultimately be akin and remained associated down to Sindarin, as the phrase Minlamad thent / estent proves, it is quite possible that this exception is a product of analogy. The initial of estent may still have been felt as a prefixed root-vowel; therefore once the latter was lowered from i to e, the initial followed, and the analogical form estent displaced the expected *istent. As for gwelwen, it is not sure that it is an old formation (note that it has no counterpart in Quenya); it may well be a late derivative of gwelw “air (as a substance)” which is the regular phonetic equivalent in Noldorin of the Quenya vilwa. If gwelwen is an old formation, its first e must be analogical of the regularly developed e in gwelw.
For U, it is difficult to say. The parallel with the A-metaphony of i to e favours the idea that it was generally restricted to penultimate syllables, but it cannot be firmly demonstrated because most u’s were lowered to o before all non-high vowels during the evolution of Sindarin, whether in penultimate syllables or not. The process has differences with A-metaphony proper and thus must be considered separately; it is presented just below.
n Sindarin and Noldorin u turns to o when the following syllable contains a non-high vowel: a, e, o and ō11). This change occurs anywhere in the word when the conditions are met.
In Noldorin at least, the final sequence -ie too lowers u to o as some infinitives testify. The o actually appears as e in the Noldorin form because of the later, fronting I-metaphony. We have no evidence for Sindarin.
Lowering is hindered by a cluster made of nasal + stop.
It does not accur either when a w or a µ follows - possibly any labial, but our evidence does not allow to be sure.
It is also prevented when two u’s follow one another.
Apparent exceptions can be explained if we consider that A-metaphony previously affected the penultimate syllabe, so that there was no longer two u’s when the conditioned lowering took place.
There is a slight problem with the name of Huan, the great hound of Valinor. It is said in the Etymologies to come from ¤khugan, so that me may expect it to show U-lowering, the evolution being ¤khugan > *χoȝan > **Hoan. Possibly vowels in hiatus were somewhat modified, so that oa shifted back to ua in Noldorin? The form is found again in Sindarin; the same explanation can hold, moreover influence of the corresponding Q word huan “hound” is likely: possibly the original Quenya name of Huan was simply transferred into Sindarin. It is also interesting that in The War of the Jewels the form Húan is found with a long ú, which would regularly remain unlowered. Obviously Tolkien slightly tried to adapt the name to evolving phonology. But this form was finally not kept in the published Silmarillion.
he vowel i affected its neighbours in various ways in the development of Sindarin. It seems that at a first stage final -i raised e and o to i and u respectively in penultimate syllabes. This is essentially the converse of A-metaphony. We have one example by Tolkien for this change: in WJ:400 he gives the evolution ¤ossī > ¤ussi in the Sindarin development of the name of Ossë. Note that the u appears as y in the Sindarin forms because of the later, fronting I-metaphony:
In Noldorin the rule is valid for the shift e > i but restricted for o > u. Here are some examples for e > i.
The shift of o > u seems to have been operative in the later monosyllables – then still disyllabic – as their plural pattern sg. o / pl. y indirectly shows (with later fronting of u > y).
But it does not sem to occur in polysyllables, at least not regularly. The plural pattern sg. o / œi (later > ei > e) in the last syllable indirectly shows that the o remained12). The alternative pattern sg. o > pl. ui (later > y) might show the shift, but it is remarkable that the u was not fronted. It is possible that this pattern is not original for o stems but borrowed from the ō < PE ā stems (variously changed to o or au later)13).
The final sequence -ia does not cause raising, on the contrary it triggers A-metaphony. Neither does -ie which can conversely lower u to o, see above. But final -io causes raising, though we have only one Noldorin example:
It also appears that final -ui caused raising: the ON pl. orkui yields N † yrchy, which points to an intermediary *urχui (with u later fronted to y).
t a second stage i fronted the back vowels a, o, u to e, œ, y respectively, in a way very reminiscent of the I-metaphony of Germanic languages. This metaphony was more widespread than the preceding one: whereas the latter were restricted to penultimate syllables, fronting occurred everywhere when i, but also ī and j stood afterwards in the word.
We can observe that fronting is caused by the rare ON pl. ending -ui too. This may appear surprising: since the first part of the diphthong is posterior it would be expected to prevent fronting. But final -ui regularly gave -y later as we see in a series of archaic Noldorin plurals (in the course of being superseded by analogic forms):
Now y is an anterior vowel – the rounded equivalent of i – so could well cause fronting. Evidently -ui was on its way to -y at the time of the Sindarin fronting metaphony. It is probable that fronting would also be caused by the Sindarin y arisen from the monophthongisation of iu, but we have no example.
The long back vowels ō and ū were not affected. Moreover, on purely phonetic grounds, we can expect that they would impede the I-metaphony of preceding vowels: as back vowels they would oppose the anticipation of fronting. We might have an example of this in the form Annui “Western” which occurs in a plural context in the King’s Letter (SD:126-9): Hîr i Mbair Annui “Lord of the Westlands”. It seems then that the plural of #annui “western” is also annui and not **ennui. The adjectival ending -ui comes from -ōja, -ūja (VT42:10), and the ū may have prevented the shift of the preceding a to e. An interesting consequence would be that all -ui adjectives would be invariable. It is unfortunate that we do not have other examples to confirm this theory. Long ō and ū were sometimes shortened later, that is why we see instances of o and u resisting fronting in Sindarin / Noldorin:
There are a few other cases where fronting is not seen though it could be expected. Extra-phonetic explanations must be sought in the scope of our model.
he plural patterns of Sindarin often show an i or e-diphthong in the last syllable, a development that must be linked to the evolution of the old -i ending. It seems that during the gradual weakening of final syllables, it was reduced to a non-syllabic i. This development is implied in WJ:400 by Tolkien’s reconstruction of the name of Ossë: ¤ossī > ¤ussi > ¤yssi (the non-syllabic character of i is marked by a subscribed crescent in the original text). Probably this sound had a quality near to [j] and palatalised the preceding consonant; a consequence was that a transitional glide in i developed between this consonant and the preceding vowel, except when the latter was already close to i: this includes i, ī, y (the rounded counterpart of i) and i-diphthongs. Eventually this epenthetic glide formed a full diphthong with the preceding vowel which was consequently broken. It is also observed when the word ends in -ia, whence we can presume that it was reduced to i (or something similar) as well. On the other hand words in -ie do not display this phenomenon14).
Since there is no mention of palatalised consonants in Sindarin, the palatalisation must have been lost later. A possibility is that it was a purely allophonic phenomenon that disappeared with its trigger, i.e. when the final i that caused it was dropped altogether. Alternatively, palatalisation may have lingered a bit after the loss of final vowels; the contrast between plain and palatalised consonants would then have become phonemic for a time; often it would have marked the opposition between singular (with plain consonant) and plural (with palatalised consonant). However, palatalised consonants would have had a very skewed distribution, occuring mostly word-finally; besides, their marking of the plural would have been largely redundant, since vowel alternations caused by metaphonies already indicated it quite clearly. Thus the distinctiveness of palatalisation would have been very weak, making it a feature prone to be lost in a functionalist point of view. Indeed this eventually happened.
In Noldorin the breaking of ō gives ui.
Breaking also happened before the palatal glide j, which must have palatalised a preceding consonant too. It seems that j was preserved for a while after a prefix or at the initial of the second member of a compound while in other positions it consistently yielded i in Old Noldorin. Later however it seems that the few cases of preserved internal j became i too (see notably ON eljadme > N eilian(w) below).
In a few instances it seems that breaking was triggered by i in hiatus word-internally.
We even have a single example where a hiatic i caused breaking and then disappeared: S gleina- “to bound, enclose, limit” related to T glania- “to bound, limit”; VT42:8. The development can be plausibly reconstructed as *glania- > *glenia- > *gleinia- > gleina-16). In an editorial note (VT42:28) Carl Hostetter wonders if it is not an error for *gleinia- but says that “Tolkien was exceedingly careful to correct errors in his citation of elements from his languages, so if gleina- here is an uncorrected error, it would be most unusual. Moreover, the phonological development of some of the Telerin and Sindarin ordinals ending in -ya discussed at the end of the appendix of this essay suggests that T glania- and S gleina- may be plausible cognates: cf. T nelya ‘third’, (archaic S.?) *neil(a), S neil, nail”. Yet this exemplifies the development finally, not medially, and the same essay has the S form seidia- < satya- (VT42:20, see just above). The Sindarin gleina- thus remains quite difficult.
But there are many counter-examples both in Noldorin and Sindarin :
Clearly epenthesis triggered by hiatic i is very much the exception. The fact that it is optional in the Noldorin words for “play” may point towards dialectal variations. It might also apply to seidia- and the curious gleina-, though in this instance it might as well be an instance of Tolkien changing his mind. It is interesting that a similar variation regarding epenthesis is found in the Brythonic languages between Welsh on the one hand, Cornish on the other hand. For instance, the plural of mab “son” is meibion in Welsh but mebyon in Cornish17).
We noticed earlier that the ultimate I-affection of a could be e, seemingly in words of a peculiar phonetic shape. In Sindarin this is seen before consonant clusters made of r + consonants, and the group ng. In Noldorin it would seem that any consonant cluster opposes breaking. There are two possibilities:
We do not observe restrictions on breaking with other vowels, but this may well be by chance. The breaking of long vowels cannot be observed before consonant clusters since long vowels are generally shortened in this position in Eldarin languages18). As for the breaking of œ, we have few examples of it anyway, and none is before a consonant cluster19).
There is a notable and embarrassing exception to breaking: the name of Elbereth, the etymology of which is given in MR:387: ¤elenbarathi. We would expect this to yield **Elberaith as *galadi gave gelaidh, yet it is not so. We must look for an explanation outside of phonology. The -i is here a feminine ending also seen in Quenya, where for instance we encounter quendi “elf-woman” (MR:229) beside the much more common plural Quendi “Elves” (S:363, WJ:361, 372). Feminines in -i do not seem to have been very common, as far as we can judge from Quenya, perhaps because of the awkward homophony with the plural ending. It must have been all the truer for archaic Sindarin since -i was almost universal as a plural mark20). It may be then that in the development of the name of Elbereth, the feminine ending, probably no longer productive, was mistaken with the plural ending, felt inconvenient in a personal name and removed or replaced by another vowel. The phonetic structure demands that, in our model, this happened between raising and breaking.
Interestingly, there is no insuperable phonetic problem in Noldorin: LRW:351 gives ON Barathi(l), N Berethil and Elbereth. The plural patterns of Noldorin show that -aCi > -eiC that can be monophthongised to -eC in unstressed syllables, which would be the case in Elbereth. Moreover, influence from the variant Berethil is not unlikely. In an external point of view, of course, the Noldorin name must simply have endured in Sindarin; but its derivation needed to be reinterpreted in light of Tolkien’s later conceptions about the historical phonology of the language.
t is interesting, if not very surprising, to find many similarities between the Sindarin and Welsh vowel affection patterns. Tolkien did not conceal his chief source of inspiration for that language. And indeed the parallels are many. Welsh has a dichotomy between final and internal I-affection. Its patterns of vowel affection are the following (Thorne, 1993).
|A-affection|| y > e
w > o
i > ai (one word)
|Final I-affection|| a > ei, ai or y (e in penult)
e > y
ae > ai
o > y (e in penult)
w > y
oe > wy
aw > y
|Internal I-affection|| a > e, ei
e > ei
ae > ei, ey
aw > ew
Historically, metaphony affected etymologically short vowels only in the development of Welsh (Jackson, 1953). Similarly, in Sindarin, only short vowel can change their quality by metaphony; long vowels can only be broken in final syllables.
However, on a phonetic point of view, Sindarin differs somewhat from Welsh when I-affection is considered. Welsh displays quite a strong tendency to assimilate sounds in final syllables to its y - a variety of i21) – and does not display Sindarin’s typical front rounded vowels y, œ22). These are much more like the products of I-affection in Germanic languages. Here are general patterns of vowel affection found in some ancient Germanic languages (Robinson, 1992; Mossé, 1950; Jolivet and Mossé, 1972).
| Old Norse|
(Old Icelandic forms)
| Old English|
(Early West Saxon)
|Middle High German|
|A-affection|| i > e|
u > o
| i > e|
u > o
| i > e
u > o
|I-affection|| a > e|
o > ø
u > y
á > æ23)
ó > œ24)
ú > ý
au > ey
| æ > e|
a > æ, e
o > œ (early) > e
u > y
á > ǽ
ó > long œ (early) > é
ú > ý
ea > ie
io, eo > ie
éa > íe
ío, éo > íe
| a > e
o > ö
u > ü
â > ae [ε:]25)
ô > oe [ø:]26)
û > iu [y:]27)
ou > öu
uo > üe
|U-affection|| a > ö28)|
e > ø
i > y
| æ > ea|
e > eo
i > io, eo
The fronting of back vowels is quite like the kind observed in Sindarin. There is a specially striking parallel between the Old English unrounding of œ to e and the same phenomenon observed in Sindarin and Noldorin. This parallel extends further: we know from the Appendix E of The Lord of the Rings that in Gondor, y too was unrounded to yield i. This is also what happened to the Old English y. In Middle English it was unrounded sooner or later in every dialect; the resulting sound depended on the dialect and the phonetic surrounding, but was frequently i. Modern English traces of this process are found for instance in kiss < cyssan and fill < fyllan.
Sindarin and Noldorin patterns of vowel affection thus appear as a kind of mix between Welsh and Germanic – sometimes specifically English – inspirations. To use one of Tolkien’s favourite metaphors, to cook his Sindarin he certainly poured much Welsh in his linguistic cauldron, but we can also track back many other bits and pieces that, blended, give Grey-Elven its unique flavour. The external and personal significance of the fact to Tolkien remains to be investigated, but may remind us his great interest in the interplay of languages on England’s soil, as we find it powerfully and disconcertingly expressed in his essay English and Welsh.
The reader is referred to the list of resources of the Elvish Linguistic Fellowship: