For many years, students of Sindarin had to piece together the system Tolkien used for grammatical mutations from disparate sources. While it was well-known that Sindarin used both vocalic mutation (aka. soft mutation or lenition) and nasal mutation, we had no comprehensive description of those mutations from Tolkien himself from the period around the publication of The Lord of the Rings (1950s and 60s). This changed with the publication of Parma Eldalamberon #23 (PE23) in September of 2024, which included a document labeled the Common Eldarin Article (PE23/133-140), hereafter designated CEA. This document gave a fairly comprehensive outline of the system of Sindarin vocalic and nasal mutations, in conjunction with a description of the Sindarin definite article. This definite article triggers vocalic mutations with singular nouns, and nasal mutations with plural nouns.
However, the definite article presented in CEA is different from the one previously known from Tolkien’s earlier writings. In particular, it takes the form e(n) before singular nouns and i(n) before plural nouns. Prior to the publication of PE23, it was widely assumed that the Sindarin singular definite article was i, as it is in Quenya, with a plural variant in. There is a separate article-like element e(n) appearing in numerous names and occasional phrases, but prior to the publication of CEA this element was assumed to have a different function, namely as a genitive marker, as described by Tolkien himself on PE17/97.
This begs the question of how much of the information in CEA builds on Tolkien’s prior work, and how much of it is new innovations. This essay compares the mutational system of CEA to Tolkien’s prior works to explore what seems to differ and what seems to be the same. In particular we will examine six topics:
Note that in the discussion below, I use > (and <) to represent historical phonological development but → (and ←) to represent grammatical sound mutations. I also use ✶ to represent primitive forms given by Tolkien but * to represent reconstructed forms (ancient or modern) not appearing in Tolkien’s own writings. Items in braces ({}) indicate deletions on Tolkien’s part, and >> or << indicate the direction of revisions.
Editorial additions to quotes are given in brackets ([]), and those additions are usually from the original editor (in most cases Christopher Gilson) unless otherwise specified. Finally, my citations in this essay of Tolkien’s original work frequently introduce editorial quote marks (“”) around glosses to facilitate reading, along with other minor editorial changes to punctuation and capitalization. Those minor editorial changes are my own and are made without comment.
For simplicity, this essay treats Gnomish of the 1910s, Noldorin of the 1920s-40s and Sindarin of the 1950s-60s as part of a single continuum of conceptual development. As such, it mostly labels linguistic constructs by date (1910s, 1930s, 1950s) rather than by language (G[nomish], N[oldorin], S[indarin]).
In CEA, Tolkien indicated that Common Eldarin had two ancient definite articles, a simple article ✶i and an “emphatic” article ✶inā. The simple article could show number as dual ✶it(a) and plural ✶im, though Tolkien stated these dual/plural forms might instead have been Ancient Telerin innovations (implying they were not used in Quenya). In Old Telerin, emphatic ina gradually displaced simple i, and the Sindarin article ending up as a blending of these more ancient forms. The result is a singular definite article e(n), with e before consonants and en before vowels. The plural article is likewise i(n), with i before consonants and in before vowels (PE23/135).
In the singular there was a variant form i, originally retained only before nouns beginning with ge- to avoid awkward combinations like e-e after the g was lost due to vocalic mutation. The example Tolkien gave was i-êl “the weeping”, presumably from otherwise unattested *gêl “weeping”, which was thus distinct from en-êl “the star”; this example appears in rough notes labeled “The anaphoric particle” that was either a follow up to or an early draft of the main CEA document (PE23/141). In the main CEA document itself, Tolkien included a difficult-to-read note that seems to say “i used orig[inally] only before e < ge” (PE23/135).
The singular definite article e(n) is the most surprising feature of the system presented in CEA. There are quite a few examples in Tolkien’s earlier writings of a singular definite article i before both vowels and consonants. Examples include:
The above examples are strong evidence that the singular definite article e· was not the norm in Sindarin as Tolkien conceived of it while writing The Lord of the Rings. It seems likely that it was an innovation of the late 1960s, probably in the CEA document itself. Tolkien’s motivations for this change are uncertain, but may have been a desire to simplify Sindarin by merging the ordinary definite article i(n) with the genitive article e(n), as discussed later in this essay.
However, something like the “emphatic article” ✶inā does date back to at least 1948, where ina was mentioned in the first version of Quenya Personal Pronouns (QPP1: PE23/86) and the accompanying document on Demonstrative, Relative, and Correlative Stems (DCR: PE23/105). In these documents this variant definite article was a Quenya emphatic or adjectival form of the ordinary definite article i, and ina was glossed “the fact (that)” or “the one that” (PE23/86 and PE23/105 respectively). Thus a primitive form *inā may have been connected to Noldorin/Sindarin genitive article e(n) as it was originally conceived in the late 1940s or early 1950s. This is a proposition I will argue for later in this essay.
In CEA the singular definite article e induces vocalic mutations before nouns beginning with a consonant, and Tolkien outlines a comprehensive system of vocalic mutations for Sindarin using this definite article as an example. These sound changes can be compared to earlier systems of mutation, such as those given in the 1930s Comparative Tables (CT: PE19/18-21), the 1920s Early Noldorin Grammar (ENG: PE13/120-121), and the 1910s Gnomish Grammar (GG: PE11/7-8).
The vocalic mutations of voiceless and voiced stops in CEA are consistent with earlier examples of mutations, dating all the way back to the Gnomish Grammar (GG) of the 1910s. That is not particularly surprising, since Welsh (Tolkien’s main inspiration for Sindarin sound mutations) shows the same set of mutations. The 1930s, 1920s, and 1910s mutations in this and later sections are taken from Comparative Tables (CT), Early Noldorin Grammar (ENG), and Gnomish Grammar (GG) respectively. Sindarin mutations from the 1950s and 60s are listed first and pulled from various sources, though they are not always examples of mutations from a definite article.
The vocalic mutations of voiceless spirants and sibilants in CEA are likewise mostly consistent with earlier mutations. Exceptions are that h was not mutated in the 1920s (though it was in the 1910s), and s was not mutated in the 1910s-20s.
The voiced liquid and nasal mutations in CEA are also mostly compatible with earlier examples. Exceptions are that initial liquids in 1930s were only voiceless (see below), and m in 1910s-20s did not mutate.
One interesting new tidbit in CEA, however, is that in modern Sindarin m did not always mutate to avoid confusion with mutations of words beginning with b (PE23/140). The examples Tolkien gave were adjectives morn “black” and born “hot”, so that Orod Vorn could be either “Black Mountain” or “Hot Mountain”, and the nouns malt “gold” and balt “force”, so that *e·valt could be either “the gold” or “the force”. Although this note was new in CEA, there are examples of unmutated m in The Lord of the Rings. Tolkien called out the specific example Imloth Melui “Sweet Flower-valley” in CEA (see also: LotR/866; VT42/18), but there are other examples as well such as Emyn Muil (LotR/373) or Ered Mithrin (LotR/1064).
It is not clear whether the sporadic non-mutation of m in modern Sindarin was always Tolkien’s intent, or whether he was rationalizing conceptual variations or errors in the way he handled m-mutations in The Lord of the Rings.
CEA gives the first clear explanation of the mutations of voiceless lh, rh in Sindarin among currently published materials. In particular, he stated that their archaic mutations were to thl, thr which were the expected medial sound developments, but in modern Sindarin these mutations were dropped and lh, rh no longer mutated.
This was not, however, the only time such changes were mentioned in Tolkien’s writings. There is a somewhat cryptic discussion in Quenya Notes (QN) from 1957 of the mutations associated with na “to” (vocalic) vs. na(n) “with” (nasal) in which Tolkien said:
[S.] na (< nā), to, towards, of space/time. with vocalic mutation. before vowel n’ ... S na, before vowels nan (nasal mutation), means “with” in sense of possessing, provided with, especially of characteristic feature. Orod na Thôn “Mount of the Pine Tree(s)”. na “to” and na “with” are therefore distinct before vowels and b, d, g, p, t, c, m, s but same before h, f, þ, r (rh), l (lh). Late forms as nan-h as for vowels, archaic nath-r, nath-l for nan-rh, nan-lh. nan|sr > nassr > nathr (PE17/147).
While this note is not clearly written, it does indicate that the mutations nath-r, nath-l were archaic, implying they were later abandoned, leaving nan-rh, nan-lh as the likely modern forms. The examples given here are nasal mutations, but in the same note Tolkien stated that the vocalic and nasal mutations of rh, lh were the same.
Note that in the 1930s, voiceless liquids lh, rh mutated to voiced l, r (PE19/19). This is because they are the historical result of ancient voiced initial ✶l, ✶r, which were unvoiced to rh, lh in initial position. The 1930s vocalic mutation simply restored the original voiced sounds, identical the developments of Welsh. Initial voiceless liquids were not mentioned in either the Gnomish Grammar (GG) of the 1910s or the Early Noldorin Grammar (ENG) of the 1920s, since these sounds were either rare or not a feature of the languages at this conceptual stage. These sounds only became common in Early Noldorin Word-lists of the 1920s compiled after ENG was written (PE13/148, 149, 152, 163, 165).
Tolkien seems to have gradually abandoned the unvoicing of initial liquids in Sindarin as he was writing The Lord of the Rings in the 1940s, but some remnants of it can be seen in names like S. Lhûn (LotR/1134) and Q. rómen vs. S. rhûn (LotR/1123), for which Tolkien was forced to contrive later explanations.
CEA also explains the Sindarin mutations of voiceless approximants, a rare mutation where Tolkien seems not to have been entirely certain of the result.
In the case of ancient ✶sy, this had become initial h- in modern Sindarin and so presumably fell together with h- > ✶kh-, with modern mutation h → ch. In the case of ancient ✶sw, this had become initial voiceless wh in modern Sindarin, but Tolkien was unsure of its mutations, considering wh → chw/cw/wh as options. He first wrote e·chwest, deleted it and replaced it with e·cwest, then below this wrote e·whest. He then wrote a lengthy note exploring the possible phonetic developments:
{This business about sw > χw, ƕ etc. is a totally unnecessary complication (following Welsh). Far better to make sw > ƕ > f in T., S. (North & South) initially. In T. su̯, sı̯ remained medially. In S. su̯ and sı̯ > ... [deleted and revised to >>]}
This business about sw is too complicated (and unnecessary). In S. sy–, sw– initially > voiceless single ꜧ, ƕ; but later ꜧ > χ̑ (ich-laut) > χ in absolute initial position, while ƕ remained unchanged (except by change > f in North Sindarin). Medially the development was to ꜧ, ƕ also but these were not intensified medially, and at time when h was lost medially became ı̯, u̯. So aswā > aƕa > ahw > āw, ası̯ā > aꜧa {> aꜧ > ai > ae} > ahy later[?] mutated[?] > ei > ai. In T. sw, sı̯ > f, s, palatalization given up as in kj etc. (PE23/137 note #10).
This note implies that intervocalic ✶sy, sw > ꜧ, ƕ > ı̯, u̯ at the same time that intervocalic h was lost, resulting in the formation of diphthongs with the preceding vowel. However, this sheds little light on what the vocalic mutation of initial hw should be. The proposed medial development seems to produce u̯ or w but that doesn’t match any of the forms Tolkien was considering for vocalic mutation.
The only other discussion I’ve found of similar sounds changes is from the Comparative Tables (CT) from the 1930s, where primitive initial ✶sj > ꜧ > χ > h with a mutated form ch, and primitive initial ✶sw > ƕ > f with a mutated form chw (PE19/21). Tolkien was still considering ✶sw > f for Sindarin in the late 1960s, since this was a [rejected] development in the CEA note given above: “Far better to make sw > ƕ > f in T., S. (North & South) initially” (PE23/137 note #10).
The vocalic mutations of ancient nasalized stops in CEA were ambiguous. In Tolkien’s prior writings, there were specialized mutations for words whose initial consonants were derived from ancient nasalized stops. In unmutated Sindarin words these became ✶mb- > b-, ✶nd- > d-, ✶ñg- > g-. These words could be detected by comparison with Quenya, where the phonetic developments resulted in a nasal rather than a voiced stop: ✶mb- > m-, ✶nd- > n-, ✶ñg- > ñ-.
In grammars and phonetic descriptions from the 1910s-30s, the vocalic mutation of such words were to a nasal (or nasalized stop) rather than a voiced spirant:
Analysis of these forms is complicated by Tolkien’s frequent representation of Sindarin initial ñ [ŋ] as ng, so it isn’t always possible to tell whether initial ng is intended to be [ŋ] or [ŋg]. Despite this, it seems that in the 1910s the vocalic mutations of words beginning with ancient nasal-stops restored these ancient clusters: mb-, nd-, ng- > b-, d-, g- → mb-, nd-, ñg-. In the 1920s the mutations were simple nasals: mb-, nd-, ng- > b-, d-, g- → m-, n-, ñ- (ng). In the Comparative Tables of the 1930s, it seems the vocalic mutations partially restored the ancient clusters, sometimes as a long nasal: mb-, nd-, ng- > b-, d-, g- → mm-, nd-, ng-.
Similar specialized mutations can be seen in Sindarin of the 1950s:
There is one possible example of the vocalic mutations of nasalized stops in CEA itself: e·alad from S. galad “radiance” (Q. ñalta). In notes from the mid-1960s Tolkien derived this word from √ÑGAL “gleam, sheen” in an attempted to reconcile the names Galadriel and Gil-galad; a root-form √ÑGAL was necessary to explain why the interior g did not vanish in the compound Gil-galad, surviving because it was originally part of a cluster l-ñg (PE17/59).
However, in the Shibboleth of Feanor from 1968, Tolkien indicated the word galad was derived from √ÑAL while still including it as the second element of Gil-galad “Star of Radiance” (PM/347). Indeed, previously in CEA Tolkien stated that “ñ– had in S. become very early ʒ > g–, and so later fell with original g and became ʒ > nil” (PE23/135). Thus it seems likely that the example e·alad “the radiance” in fact represents the mutation of ancient simple ✶ñ- > g-, with a result identical to that of original initial g- as in e·alað “the tree”. Either Tolkien forget his original need to explain the survival of interior g in Gil-galad or he contrived a new explanation, perhaps as a juxtapositional genitive meaning “star of radiance” rather than a true compound; see the section on the Genitive Article below for further discussion Sindarin genitive constructs.
Assuming the above is true, then CEA contains no examples indicating what the vocalic mutations were for ancient nasalized stops.
The mutation of gw was given separately in CEA since its phonological origin was distinctive. Ordinary stop g → ’ due to the loss of g between vowels, but labialized stop gw → w because the phonetic development w > gw occurred only at the beginning of words and not in the interior. Thus the mutation of gw → w was the result of an inhibited initial sound change rather than a reflection of an interior sound change. This mutation, however, was an old one dating back to Gnomish of the 1910s:
In CEA the plural definite article i(n) induces nasal mutations before nouns beginning with a consonant. There is a complication with nasal mutations not seen in vocalic mutations, however. In particular, the n in the plural definite article not only mutates the following sound, but can itself be modified as part of the mutation. Early in the CEA document Tolkien said that the plural article was i before consonants and in before vowels, but in the actual examples sometimes the n vanishes and other times is retained or modified.
Because of this more complex outcome, Tolkien also explored the historical phonological developments of nasals combined with other consonants, with the nasal mutations mirroring (but not always precisely matching) these phonological developments. In the analysis below, I include both these phonological developments as well as the resulting nasal-mutated forms that Tolkien presented as the end result.
Also note that both in CEA and before CEA, Tolkien’s system for Sindarin nasal mutation is not identical to Welsh’s system of nasal mutations. In particular, while voiced and voiceless stops become voiced and voiceless nasals respectively in Welsh nasal mutation, Sindarin voiceless stops become voiceless spirants instead. This difference between Sindarin and Welsh seems very important to Tolkien, and he strived to maintain it throughout all of his works. The phonological developments leading to this specific mutation will be discussed in a section of its own later in this essay.
In CEA, the nasal mutations of voiceless stops were voiceless spirants, and the n in the definite article survived. This is in contrast to Tolkien’s earlier writings, where the n generally vanished.
Thus in CEA the nasal was retained before mutated voiceless stops, but prior to CEA the nasal was lost. This loss of n dates all the way back to the Early Noldorin Grammar (ENG) of the 1920s: p → i·f, archaic i·mf; t → i·th, archaic i·nth; c → i·ch, archaic i·nch (PE13/121). In CEA, Tolkien justified the retention of the n before mutated voiceless stops as follows:
The spirantalizing of stops after a nasal was however very much later: a purely S. change only affecting the medial stops and not those that had become final after the loss of all original CE final vowels (–mp, –nt, –ñk remained); the nasal was therefore not phonetically lost before the new spirant (PE23/137).
In this conception of the phonetic development, medial nasal + voiceless stops became spirants after final vowels where lost, so the change did not apply to final combinations. Tolkien may have introduced this change to make it easier to distinguish nasal mutations of plural words beginning with voiceless stops versus voiceless spirants (as described in the next section), such as *in·thyss ← toss “bush” versus *i·thyss ← thoss “fear”.
Tolkien contrasted the nasal mutations of voiceless stops with those of ancient aspirates that had become voiceless spirants, where the nasal was lost earlier. This n-loss before original voiceless spirants is consistent with Tolkien’s writing elsewhere.
In CEA, Tolkien justified the loss of the n before original voiceless spirants as follows:
The first change was one common to T. and S. and so very ancient. After the loosening of the aspirates to spirants a preceding nasal was lost, and the vowel before the nasal was nasalized and lengthened. The nasality was later lost (in both T. and S.). The lengthened vowel fell in with the later development of original long vowels, but was shortened in S. in the unstressed proclitic article. So that in result the plural article in > i, leaving the initial consonant unaltered (PE23/137).
Thus it seems Tolkien had two distinct stages for the sound changes of n combined with voiceless spirants:
In CEA, the nasal mutations of voiced stops were voiced nasals, along with the loss of the n from the plural article in. This is consistent with nasal mutations seen prior to CEA.
Tolkien noted that in spelling these mutations were often represented as nasal-stop clusters:
... for i·mair, i·nuir written forms were sometimes i·mbair, i·nduir to distinguish such cases from those with original initial m, n; they were not needed in the case of nouns beginning with g–, since no words had radical initial ñ in Sindarin, but nonetheless i·ñuin, the stones, was sometimes spelt i·ñguin (PE23/139).
As an authorial note to the above he said:
These spellings descend from a period when mb, nd, ñg medially still remained and had not yet become mm, nn, ññ. Possibly all three still remained before the main word stress, sc. after proclitics. This change took place medially later when followed by a main word-stress ... The nasalization of the stops b, d, g was not complete at the time when the Noldor first became acquainted with Sindarin: mb was the first to change, and had in all positions > mm; nd had become nn medially, but remained finally (after the loss of older final vowels); ñg was still unchanged. Later developments were: –nd > nn > n finally in unstressed syllables, and eventually in all positions; ñg finally > ññ > ñ, and also initially (as the mutation of g–) (PE23/140).
Thus i·mair, i·nuir, i·ñuin were sometimes spelled i·mbair, i·nduir, i·ñguin. The pronunciation was the result of shortening of long nasals at the beginning and ends of words, but these sound change had not yet occurred at the point where the Noldor encountered Sindarin, and the spelling was codified early and not always updated when the pronunciation changed.
Also possibly relevant is that two of Tolkien’s examples, bâr “dwelling” and dôr “land”, are normally represented as being derived from ancient roots beginning with nasalized-stops, √MBAR and √NDOR respectively. This may imply that such nouns had the same nasal mutations as those with ordinary voiced stops, probably via reductions of overlong nasalized stops: in + mb, nd, ng > immb, innd, inng > imb, ind, ing > imm, inn, ing > i·m, i·n, i·ñ. The orthographic ambiguity mentioned above would make it hard to tell from the written forms.
Tolkien’s earlier writings may have had similar orthographic ambiguity, which makes it tricky to determine whether or not there were any specialized nasal mutations for ancient initial nasalized stops in Tolkien’s early writings (see the sections on the Genitive Article and Mixed Mutation below). In the Early Noldorin Grammar (ENG) of the 1920s, there were no such specialized nasal mutations: both ordinary voiced stops and ancient nasalized voiced stops mutated to simple nasals: b, mb → i·m; d, nd → i·n; g, ng → i·ng (PE13/120), the last presumably ñ spelled as ng.
In CEA, nasal sounds were not modified in nasal mutation, but the n of the plural definite article was lost. This is consistent with the nasal-shortening discussed above for the mutations of voiced stops, and is also seen in earlier writings:
The velar nasal ñ was a special case in CEA. This nasal had early become g- in initial position, so that this combination fell together with the nasal mutation of original ancient g.
For liquids Tolkien outlined the phonetic developments but gave no example words (PE23/137):
There are also no examples of these nasal mutations for plurals in Tolkien’s earlier writings, though there are words with the relevant phonetic developments in the interior of words: [rejected] caralluin < caran + luin (VT48/30) and Caradhras < caran + ras (LotR/1113; PE17/36). Lack of examples makes it difficult to determine how exactly these mutations function. Probably the ll shortened (analogous to the shortening of double-nasals and ss) so that l → i·l. In the case of r, it might have remained r → i·ðr or ið·r, or it may have simplified to r → i·r like the simplification of i·thr to i·rh (see below).
However, there are the (previously mentioned) Quenya Notes (QN) from 1957 where Tolkien indicated mutated n·r > ð·r likely survived. In particular, in the discussion of the nasal mutations of nan “with” Tolkien said: “nað before r (nan-r > nađr)” (PE17/147). Unlike nath-r and nath-l from the same note, he gave no indication that nað-r was archaic.
Tolkien described the nasal mutations of voiceless sibilants and approximants in CEA in some detail:
The mutation of s was the result of the sound change whereby ns > ss, with the double-s then shortening at the beginning of words (analogous to double-nasals). It seems h < sy fell together with h < kh with same nasal mutations (vocalic mutations were likewise identical, see above). Tolkien was uncertain of the vocalic mutation of wh, but there is no indication of similar uncertainty for its nasal mutation; perhaps the early presence of n inhibited the complex mutations associated with isolated sw > wh.
Tolkien gave a detailed description of the nasal mutations of voiceless liquids in CEA:
There are no example words with these nasal mutations in Tolkien’s earlier writings, but as discussed above under vocalic mutations, Tolkien did address the sound changes themselves in Quenya Notes (QN) from 1957, where he compared the (vocalic) mutations of na “to” with the (nasal) mutations na(n) “with, of”. In these 1957 notes, vocalic and nasal mutations were identical, with archaic thr/thl and (presumably) modern rh/lh: “archaic nath-r, nath-l for nan-rh, nan-lh. nan|sr > nassr > nathr” (PE17/147). In these 1957 notes, however, it seems n was restored by analogy before rh, lh: nan-rh, nan-lh.
As with vocalic mutations, Tolkien treated the nasal mutation of initial gw separately because of its distinct developments versus simple initial g:
Here the ordinary phonetic development preserved ancient nw, a favored combination in ancient Eldarin. Tolkien gave an explicit derivation: ✶im wendēi > inwendi > in·wind due to mw > nw. However after the nasal mutation of ordinary g became g → ñ (written ng), the nasal mutation of gw was reformed to ñw (written ngw) by analogy:
... pl. in·wind. The last form is rarely found. The usual form was even in early records i·ñwind (transcribed i·ngwind). This was due to analogy with words beginning with g–, as i·ngelaið, the trees, i·nguin, the stones (PE23/139).
The clearest example of this mutation appearing in Tolkien’s earlier writings is in Gwanur “the twins” from The Lord of the Rings appendices (LotR/1054). Tolkien explicitly called out this example in CEA:
Cf. also in LR III 335, Hauð in Gwanur “Grave-mound of the Twins”. Gwanur “twin-birth” was used as the plural (dual) of gwanon “one of a pair of twins”; but older usage would have treated it as dual pair-word with sg. article: e·Wanur ... Gwanur is derived from gwanūr < CE wo-nōrē (stem ONO, NŌ, beget) (PE23/140).
Here Gwanur is functionally a plural because it has a plural definite article, and in Gwanur is equivalent to i-Ngwanur [i-ŋwanur]. Somewhat strangely, there is another definite plural form i·Wenyn “the Twins” based on singular gwanon, appearing in The Shibboleth of Fëanor from 1968 (PM/353). Perhaps in this instance Tolkien imagined this to be the normal nasal mutation of gw, with im-w > in-w, followed by nasal loss to i-w rather than normalized to i-ñw as in CEA.
Note that the ambiguity of word-divisions in examples like in Gwanur (rather than i·ngwanur) also complicates analysis. At various points in the 1930s, 40s and 50s Tolkien wrote “the Noldor” as in-Gelydh (S/238), inGeleidh (LR/201), or i·Ngeleð (PE22/41). This variation in placement (in-g vs. ing vs. i-ng) along variations in orthographic representations like ñ vs. ng makes it more difficult to distinguish actual conceptual modifications from mere orthographic discrepancies.
As noted above, one distinction between the mutational systems of Welsh and Sindarin was the nasal mutations of voiceless stops. These became voiceless nasals in Welsh but voiceless spirants in Sindarin. This was already a feature of the earliest iteration of nasal mutation, which Tolkien introduced in the Early Noldorin Grammar (ENG) of the 1920s (PE13/120-121). In this 1920s system, the nasal n of the plural definite article survived (with some assimilation) archaically before mutated voiceless stops but was lost in the modern form of the language, for example in the nasal mutation of puig “clean”: archaic i·mfuigin, modern i·fuigin.
This continued to be the case in Noldorin of the 1930s, such as the forms i phinn “the Elves” ← penn and ithail “the feet” ← tâl (PE22/67). Tolkien had forms like i thíw “the letters” (TI/182) and i Pheriannath “the halflings” ← perian (SD/46) in Lord of the Rings drafts of the 1940s, forms that survived into the published versions (LotR/305, 768) from the mid-1950s (1st edition) and mid-1960s (2nd edition). The only possible exception I am aware from this period is mhellyn în phain “all his friend” from the King’s Letter (SD/129), where the n may have survived because it was from a pronominal element în “his” rather than a definite article. Thus the loss of n from the definite article was the norm in Sindarin up through at least the mid-1960s.
It is also well known that the medial developments of Sindarin mirror those of Welsh, in that combinations of nasals + voiceless stops developed to long voiceless nasals. This was not a feature of Gnomish of the 1910s, which generally retained mph, nth, nch medially, but clear indications of these developments can be seen in The Etymologies of the 1930s:
In the examples above, it seems the medial voiceless nasals were eventually voiced, masking the voiceless nasal stage, but medial voiceless nasals were a feature of (Old) Noldorin of the 1930s, as can be seen in the existence of dedicated tengwar for voiceless nasals, named {nhui >>} anhau and {mhui >>} amhui (PE22/30 note #98).
One notable manifestation of this sound change is uninflected vs. inflected nasal-infixed past tense forms, which clearly demonstrate the distinctions between final and medial developments. Inflected past forms from the Early Noldorin Grammar (ENG) of 1920s have voiceless spirants, at least to begin with:
The above appeared in draft notes for ENG. The inflected past of dag- was revised to dengin, and later in the final typescript Tolkien had inflected pasts like meinniog for maint (PE13/131). This indicates Tolkien introduced medial nth > nn and nch > ng sometime in the 1920s.
Inflected past forms of the 1930s consistently have long nasals, presumably originally from long voiceless nasals:
Tolkien described the Sindarin developments in detail within notes from 1962:
Similarly in oldest Sindarin mp, nt, nc were preserved. In North Sindarin they became universally mf, nth, ŋch at same time as lt, lp, lc and rt &c. > lth, lf, lch. The normal development in North Sindarin was then to restopping when final > mp, nc, lt, but nth, lf, lch, rth, rf, rch remained spirantal, medial nch > ¯ch. Later again nth > nt finally. (As in Sindarin general[ly] m, n tended to be lost before another consonant, camprû > camfrū > cāfru.)
In Oss[iriandic], West & South Sindarin changes proceeded as in North Sindarin as far as mph, nth, ŋch. Then lt, mp, nt, nc were restopped finally. Medially lþ, mph, nþ, ŋχ (though for long without change of spelling) became long voiceless l, m, n, ŋ, English transcription lh, mh, nh, ngh. In late Beleriandic Sindarin these became voiced ll, mm (m), nn, ng (= ŋg) unless followed still by a stressed syllable (PE17/131).
These developments can be seen in past participle forms from around this time:
Tolkien addressed this topic again in notes associated with The Rivers and Beacon-hills of Gondor written in 1967-69, with editorial editions (in []) from myself based on information elsewhere in this document:
The forms canthui, enchui, tolthui [“fourth”, “sixth”, “eighth”] are those of the southern Sindarin dialect adopted by the Noldor. In the Northern dialect (which perished in the course of the war against Morgoth) nt, nc, mp had remained unchanged [e.g. cantui, encui]. In the Southern dialects nt, ñk, mp remained when standing finally—or more probably the spirant was re-stopped in this position; for similarly final lth > lt, though rth remained finally. Medially however nth (nþ), nch (ñχ), mf (mp with bilabial f), and lth (lþ) became long voiceless n, ñ, m, l, though the old spelling was mostly retained (beside nh, ñh, mh, lh), and among those to whom Sindarin became a language of lore, as the men of Gondor who were or claimed to be of Númenórean race, the spirant was reintroduced from the spelling. In true Sindarin of the Elves or Elf-friends of the early ages the final form was often introduced medially (The Rivers and Beacon-hills of Gondor, 1967-69, VT42/27).
Here the examples are more nuanced. Tolkien had forms with nasal-stop, nasal-spirant, or voiceless nasal combinations. These were apparently variations between the North and South Sindarin dialects as indicated above, or perhaps reformations from spelling “among those to whom Sindarin became a language of lore”:
The last example is for liquid l + stop t combinations which had a similar set of variant forms.
All of the above begs the question: if combinations of nasals and voiceless spirants became voiceless nasals medially in the 1930s through mid-1960s, why did this not occur in the nasal mutation of voiceless stops. In other words, why did Sindarin mirror Welsh in its medial developments of voiceless nasals, but not in its nasal mutations?
Hints at an explanations can be seen in the 1962 note give above: “As in Sindarin general[ly] m, n tended to be lost before another consonant” (PE17/131). If there was some factor that resulted in the loss of the nasal from clitics like the plural definite article in, then the nasal could have lasted long enough to change the voiceless stop to a spirant, but then vanished before it could transform it to a voiceless nasal. Such a phenomenon can be seen in the Noldorin/Sindarin word ifant “aged = year-full”, a combination of în “year” and pant “full” (LR/358, 400; WJ/192). Its full derivation appeared in The Etymologies of the 1930s:
This example suggests that a nasal could be lost before a single voiceless spirant derived from a voiceless stop under some conditions, perhaps at morpheme boundaries or before the main stress of compounds. This could explain both the loss of n in the nasal mutation of voiceless stops, as well as their transformation to voiceless spirants but not voiceless nasals. David Salo proposed a similar theory in his Gateway to Sindarin (GS/49 §4.112), giving Arathorn < Aran-thorn as another possible example, which was confirmed after the publication of his book (PE17/32).
In CEA Tolkien seems to have imagined a similar nasal loss before ancient voiceless spirants derived from aspirates:
The first change was one common to T. and S. and so very ancient. After the loosening of the aspirates to spirants a preceding nasal was lost, and the vowel before the nasal was nasalized and lengthened. The nasality was later lost (PE23/137).
However, he was careful to state that this nasal-loss did not apply to nasals before voiceless spirants resulting from voiceless stops:
The spirantalizing of stops after a nasal was however very much later: a purely S. change only affecting the medial stops and not those that had become final after the loss of all original CE final vowels (–mp, –nt, –ñk remained); the nasal was therefore not phonetically lost before the new spirant (PE23/137).
The net result was that in the CEA system of mutations, the n was lost before voiceless spirants derived from aspirates, but not from voiceless spirants derived from voiceless stops. This again begs the question: why did these nasal + voiceless spirant combinations not proceed on to become voiceless nasals as they did in Welsh?
It is my belief that in CEA, Tolkien avoided voiceless nasals by removing this particular sound change from Sindarin’s phonological history. At no point in CEA did Tolkien mention the development of long voiceless nasals from nasal + voiceless stop combinations, though he did discuss the development of long voiced nasals from combinations of nasals + voiced stops (PE23/140). In this discussion, he also listed all geminate consonants: “The long consonants in Sindarin were the nasals mm, nn, ññ; ss; ll, rr; and voiceless llh (< ls, lþ)” (PE23/140). The omission of long voiceless nasals is notable, especially given the inclusion of voiceless llh.
Secondary evidence for the omission of voiceless nasals can be seen in a discussion of these sound changes and their effect on past tense forms, appearing in a rough notes with a nib pen accompanying CEA. Going by Christopher Gilson’s reading, the note states:
In Telerin the aspirates > spirants earlier than in Q.
These rarely occurred medially after consonants: the chief cases being after infixed nasal, as in past tense and other derivatives of verbal stems.
In T. [and] S. a nasal before f, þ, χ was lost with nasality, but ns had earlier > ss; hence T imph, inþ, iñχ > ī̜f, ī̜þ, ī̜χ (in case of article unstressed ī̜ > i), but in-s > iss ́. Hence mutations in S. of ph, th, kh, s > f, þ, χ, s.
Only in S. and much later medial nt, mp, ñk, at same time as r/l + t, k, p, became spirant. So pl. im tonōi > in tonoi > tonəi > tonī > in·tonĭ > in þuı̯n > in þuin, without loss of n.
Change differently in·chelair, later, as in tancher > tanker “they fixed”; tanchen “I fixed”; 3 s[g.] tanc, adanc.
Keep nch in sp[elling] S. pronounced this with nasal vowel, but nc often [?came] in from final, as adancer for adancher. But S. [?pron.] n + þ, f, χ all as [???] [???].
in khandi [>] i chend; in [?kandi] [>] in chend. [Added in ball-point:] e·chand (PE23/138 note #12).
Editorial marks ([]) in the above are from Christopher Gilson. I have omitted some of Gilson’s editorial markers of uncertainty to improve the quote’s readability, retaining only those where uncertainty was the greatest.
In the above, there are inflected past forms tancher “they fixed”, tanchen “I fixed” [not *tanger, *tangen] vs. uninflected tanc, presumably from the verb tag- “to fix” < √TAK. Compare these to manthen/maint from the 1920s vs. mennin/mant from the 1930s (see above). Somewhat cryptically Tolkien said “but nc often [?came] in from final”. This may explain tancher > tanker “they fixed”, which could be an example of restoration of the final nc rather than the historical sound developments; compare also “adancer for adancher”.
Of unclear significance is Tolkien’s note that “S. pronounced this with nasal vowel”. He may have meant the pronunciation was a̜ŋχ or ā̜χ. The end of the sentence might have clarified his intent, but it is unfortunately illegible.
Regardless of the exact pronunciation, the inflected past forms tancher or adancher (possibly revised to tancer or adancer by analogy with the uninflected past) supports the idea that medial sounds did not develop ŋch > ŋh > ŋŋ > ŋg in CEA, compared to how they did in the aforementioned notes from 1962 (PE17/131) and 1967-69 (VT42/27).
Assuming my reasoning above is correct, CEA may not be the only place in the 1950s and 1960s where Tolkien omitted voiceless nasals as developmental stage from the phonological history of Sindarin.
There is another past tense form echanthel “*you shaped” appearing in notes also from 1969 (VT47/38) which seems to indicate the retention of nasals + voiceless spirants medially. Combinations of nasals with voiceless spirants also appear in some of the variant forms of ordinal numbers (canthui “fourth”, enchui “sixth”) from the aforementioned notes associated with The Rivers and Beacon-hills of Gondor written in 1967-69 (VT42/10), but since those notes also explicitly mention voiceless nasals as a development stage of Southern Sindarin, these are more likely dialectical variants or reformations from spelling.
A more compelling example appears in Quenya Notes (QN) from 1957:
√STEN-. Since nt &c. no longer are held to yield nn &c., the word for “short” should be ✶stenna, √STEN- “cut short, limit, confine, cramp”. Q thenna/senna. S thenn, esten (PE17/185).
This note was pointed out to me by Vyacheslav Stepanov in a Discord conversation from Januarny 12, 2025 (https://discord.com/channels/397489292185960468/970002661732667464/1328033335976394853), where he suggested (and I agree) that Tolkien was likely concerned with Ann-thennath from Chapter 11 of Book One of The Lord of the Rings (LotR/193). Here it seems Tolkien believed that thent “short”, which dates back to The Etymologies of the 1930s (LR/388), was no longer suitable “since nt &c. no longer are held to yield nn &c.” This likely means its class plural form would no longer be thennath but something like *thenthath (exact form uncertain, since dissimilation of at least one th in such a situation is likely).
Assuming this reasoning is correct, retaining Ann-thenneth required a change of the uninflected form to thenn < ✶stenna. This makes perfect sense if, in that moment, Tolkien imagined Sindarin’s phonological developments to be nt > nth but not nth > nh > nn. Thus it seems likely that Tolkien was toying with the notion of abandoning the long voiceless nasal phonological stage as early as 1957, though clearly he changed his mind since such a stage was restored in the aforementioned notes from 1962 (PE17/131).
Summarizing the above, here is what I believe were Tolkien’s thoughts on voiceless nasals as a phonological stage at various points in his life, using nt as an example:
As an additional note, it seems Tolkien also imagined there was restopping of final -nth > -nt at all points, except that (a) in The Rivers and Beacon-hills of Gondor (1967-69) this restopping did not occur in North Sindarin because nt was unchanged in all positions, and (b) in CEA (1969) this restopping occurred only medially for North Sindarin, since final -nt survived in all dialects.
Similar phonologic developments applied to mp and ŋk, with the addendum that ŋh > ŋŋ > ŋg in those circumstances where Tolkien imagined that long voiceless nasals were voiced (e.g. the 1930s and 1962).
Prior to CEA, the singular definite article was i· instead of e·, but even in this period e· (and en·) appeared in phrases such as Taur e-Ndaedelos “Forest of the Great Fear” (LotR/1134) and Conin en Annûn “princes of the west” (LotR/953). Prior to the publication of CEA, most students of Sindarin assumed this was some kind of genitive marker. For purposes of discussion, we will refer to this and related forms as genitive articles or genitive markers (the latter when they could be indefinite), though they could be prepositional instead.
The notion of a genitive article dates all the way back to the Gnomish Grammar (GG) of the 1910s, which had ordinary article i(n), genitive article na(n) and dative article i(r), where the parenthetic value indicates the prevocalic form (PE11/9). Tolkien described the origin of the genitive article na(n) as follows [editorial additions are mine]:
root ī: This gave in the plural either ī or īn and in the genitive īn, but before words beginning with nasalized-explosives nd, mb, ng (a fairly numerous class originally) īn- also was developed in other cases. This gave rise on one hand to the double ina- genitive, plural inan- [from combination with genitive prefix a(n)·], whence the present ordinary “apocope” forms na-, na(n)- are developed, and on the other hand to a generalization of in / i forms according to new rules. Similarly with the genitive prefix a· plural an·, now general a· or an· (PE11/7).
At this stage the genitive article na(n) caused vocalic mutations:
As mentioned above, Gnomish of the 1910s also had a “genitive prefix” a·. In the Gnomish Lexicon, Tolkien seems to indicate it caused nasal mutation:
a· prefix causing initial consonant change (n· mutation), a mark of genitive employed now both with and without -a termination — also often syncopated leaving only the mutation (PE11/17).
However, actual examples had an before vowels and a before consonants with vocalic (rather than nasal) mutation:
Finally, GG had genitive suffixes -a/-n (singular post-consonantal and post-vocalic) and -ion/-thon (plural post-consonantal and post-vocalic), but these are outside the main focus of this essay.
The main functional distinction between na(n) and a(n) in the 1910s seems to be definite vs. indefinite. In particular:
The article is used much as in English in reference to what has already been mentioned, in generic usages and so on. It differs in being much more frequently used with proper names — noticeably in the genitive where with a name na(n) is almost always preferred to a(n) (PE11/9).
There is an attested example of na being used with a proper name: Tirin na Gilweth “*Tower of Gilweth” (PE13/95), with a curious lack of mutation. However, there are also examples of an with proper names, as seen in the list above.
Both na and a continued to appear as genitive markers in 1920s, for example in Dor-na-Fauglith “Land of Thirst” (LB/275) and Loth-a-ladwen “Lily of the Plain” (LB/149), though the examples are few enough that the distribution between definite/indefinite is no longer clear, possibly abandoned, or maybe even reversed.
Starting in the 1930s, Tolkien explained genitive na as a derivative of the allative root ANA/NĀ “to, towards”, as first mentioned in The Etymologies composed around 1937:
NĀ¹ ... N[oldorin] na “with, by”, prefix an-. Also used as genitive sign (LR/374).
In the 1930s and 40s the genitive marker na still took the form nan before vowels:
It generally took the form na before consonants:
Where examples of na had an associated gloss, this gloss generally omitted the English word “the”. The only counter-examples I can find is Pennas-na-Ngoelaidh “History of the Noldoli” from the 1930 title page to the Qenta Silmarillion (SM/77) and na ond “by the grey stone” from drafts of Thror’s map for the Hobbit from around 1936 (AotH/56). It therefore seems probable that by 1937 na(n) was no longer definite, and might even have been specifically indefinite.
The mutations in this period are rather inconsistent. Na sometimes caused vocalic or nasal mutation, and often no mutation at all:
The first example is clearly soft mutation of gond “stone”, but is prepositional rather than genitival. The second example might be either vocalic or nasal mutation, since the root of golodh “Noldo” was √ÑGOL. Most of the remaining examples involve voiced stops and show no mutation at all. The final example Taur-na-Danion is ambiguous, since we don’t know whether in that instance singular “pine” is Tan or Dan; later it was thaun (LR/392) > thôn (PE17/82). Similar inconsistencies for en prompted David Salo to craft his theory of mixed mutation (GS/79-80), as discussed later in this essay.
There is evidence of a genitive marker a(n) in Lord of the Rings drafts from the 1940s:
The last two examples appear to use nasal mutation. These genitival a(n) may therefore refer back to the “prefix an-” mentioned in The Etymologies of the 1930s as a counterpart to genitive marker na “with, by” (LR/374; full quote given above). Regardless, it seems likely these were holdovers from genitival a(n) from the 1910s and 20s, with the caveat that it no longer seems to be specifically indefinite based on the glosses. Indeed, it seems likely that na(n) and a(n) had flipped their indefinite/definite functions in the 1930s and 40s.
I can find little evidence of genitival a(n) in Tolkien’s texts from the 1950s. The only exception is a passing mention in a discussion of the origin of genitival en (PE17/97, discussed below). It is my opinion that Tolkien replaced a(n) with e(n) by the 1950s. It is possible that 1930s-40s genitival a(n) was crowded out by the conjunction a “and” (<< ar) and dative an “to, for”, both introduced during the writing of The Lord of the Rings.
There is an additional genitival construction in the 1930s and 1940s worth mentioning here. In this period Tolkien often put two nouns in juxtaposition to indicate a genitival relationship, with the second noun undergoing vocalic mutation as if it were an adjective.
By the 1950s, Tolkien had abandoned vocalic mutations for such juxtapositional genitives. Compare the later forms of names from the list above.
These “juxtapositional genitives” also existed in Tolkien’s earlier writings from the 1910s and 20s, sometimes with vocalic mutations and sometimes without:
In the 1930s, there are further examples of the use of definite article i(n) in what could be juxtapositional genitives, mostly from The Etymologies written around 1937:
It is possible, though, that this i(n) is instead a variant (plural and/or definite) form of the genitive marker an or na. The best evidence for this is Pennas inGeleidh “History of the Gnomes” from the 1937-8 title page of the Quenta Silmarillion, which was a revision of Pennas na-Ngoelaidh from the 1930 title page (SM/77).
It may not be a coincidence that the examples above are all plural forms. There are, however, plural genitives from the late 1930s without i(n):
Note that the English word “the” is absent from the glosses of the above examples (if they have one), so these may all be specifically indefinite. If so, the previously mentioned examples of genitival i(n) may be markers of a definite plural genitive. If so, this was likely in addition to its normal function as the plural definite article. Assuming this reasoning is correct, these 1930s uses of a genitival i(n) seem to be in competition with 1940s a(n) which also functioned as both a definite and plural genitive marker (see above).
Genitive na is mentioned again in Common Eldarin: Noun Structure (ENS) written between 1951-1952, still tied to the root ANA/NĀ:
The influence of the adverbial/prepositional stem ANA/NA has also been assumed [as the basis for Quenya allative/dative n/nna], but the sense and functions of ANA are not really very close. It is true that in Q. na appears with sense “to”, but this at nearest means “towards, to a position near, alongside”; whereas the allative/dative n/nna means “up to, to, at (arriving at the point)”. The original sense of Eldarin ana was plainly “at side of, alongside, besides”, hence also “moreover, in addition, plus” (seen in use of an- as an intensive prefix), and so an or na in some languages has the sense “along with, with, accompanied by, provided with, associated with” and the like. Cf. Bel. [= Sindarin] na which forms virtually adjectival expressions: as Taur na Foen “The Forest of Foen” i.e. which included the mountain called the Foen (PE21/79).
Another explanation appeared in a letter to David Masson written in 1955:
Orod-na-Thôn ... na is rather a multi-functional word in appearance! Its functions in Quenya/Sindarin can however probably be derived from ANA/NĀ “allative” base. In Sindarin it is a preposition and in na-Thôn functions like French à “provided with, marked by, with” etc. (PE17/82).
Tolkien undertook an extensive examination of the Quenya and Sindarin derivatives of ANA/NĀ in Quenya Notes (QN) written in 1957. The excerpts below focus on Sindarin, including editorial additions from Christopher Gilson indicated by brackets [] and Tolkien’s deletions by braces {}. Tolkien first wrote a rough draft which he rejected with a vertical stroke:
AN/NĀ “to, towards”. S an “to”, na. an chiefly in forming datives of pronouns, as anim “to myself”, ammen “to us”... Probably distinct is S na “with, provided with”.
S na- in na-chaered, Orod na Thô[n]. na also had sense “towards, to, added to”.S an “to (dative)”. na, spirant [mutation] is for nat-, nad- “towards place”. na-chaered. nā. {na-} nan-. ana.
S an, vocal[ic mutation] (ana) “dative with place or person”.
na, nasal [mutation] < nan.
na, {voc.} n- < nā “with, provided with, with addition” (PE17/147).
In this rejected note, Tolkien seems to be trying to determine the mutations for an “to” (dative), na “with” (genitive), and na “towards” (allative). He considered vocalic, nasal or spirant mutation (AKA stop mutation), which might be distributed in various ways between among the prepositions under examination. He rewrote the note in full, with relevant Sindarin excerpts given below:
√ANA/NĀ, “to, towards; added to, plu-” ...
S an, dative chiefly with pronouns or persons. < ana, hence vocalic mutation, {but preserves g, d} but takes form m before m, b. “to, for”. dative after word give ... S {ónen an >> ōn anim >>} ōn annin ...
na (< nā), “to, towards, of space/time” with vocalic mutation, before vowel n’.
{[Deleted:] but also = “with, provided with, characterized by possession of”. In first sense it takes analogical (? in the area < nā na·) nasal mutation, or later vocalic}
S na, before vowels nan (nasal mutation), means “with” in sense of possessing, provided with, especially of characteristic feature. Orod na Thôn, Mount of the Pine Tree(s). na “to” and na “with” are therefore distinct before vowels and b, d, g, p, t, c, m, s but same before h, f, þ, r (rh), l (lh). Late forms as nan-h as for vowels, archaic nath-r, nath-l for nan-rh, nan-lh. nan|sr > nassr > nathr. nað before r (nan-r > nađr). For this na(n) Quenya used suffix -va. S i-arben na megil and “The Knight of the Long Sword” = Q arquen andamakilwa. Thus after arose genitive, as Aran lintaciryaliva, S ara cîr lim or aran na chir lim = Aran linta ciryalion (PE17/147).
In this revised version, he seemed to settle on na “towards” having vocalic mutation, and na(n) “with” having nasal mutation. The mutations for dative an “to, for” will be discussed in the final section of this essay, covering mixed mutation. We can compare this paradigm to attested mutations from the 1950s and 60s, na “towards” being the easiest since there are only two examples:
The first example is clearly na “to(wards)”, but the mutation could be either vocalic or nasal. The second example is clearly vocalic mutation. The meaning “at” is not obviously allative, but does fit the allative sense “at (arriving at the point)” seen in the quote from Common Eldarin: Noun Structure (ENS) given above (PE21/79). Therefore, both these examples are consistent with vocalic mutation for allative na.
Attestations of na “with, by, of” are more ambiguous. Most involve no mutation at all (Taur-na-Neldor) or ambiguous mutations (Taur-na-Chardhîn). I can find only three examples that should be unambiguously mutated, and they are unfortunately contradictory:
The first example demonstrates the nasal mutation explicitly described in QN where it appears. In the second example, n’ is elided before Arðon which would seem imply vocalic mutation. The third example either has no mutation or is something exotic like mixed mutation, otherwise the initial d would become dh (vocalic mutation) or n (nasal mutation).
1957 QN indicates the pre-vocalic form is nan, as in nan ellas (PE17/97). However, our only other attested example n’Arðon from 1968 has elision instead (PM/348). It is possible that the unmutated form of Arðon is *garðon, but the loss of g still points to vocalic mutation instead of nasal mutation.
Glosses indicating definiteness from this period as also inconsistent. Some glosses have English “the” while other glosses omit it:
Finally, there was what seems to be a (definite?) plural variant nia that appeared in a couple places:
The second example was revised {nia Chîn >>} i Chîn, which might be the plural form of e(n) as discussed below.
By the 1950s Tolkien had introduced a new genitive marker e(n) along side na(n), both of which remained as genitive markers throughout his later writings up until the change of e(n) to a definite article in CEA. The pre-CEA distinction between en and na isn’t clear. It has been hypothesized that en is a definite variant of na. However, as noted above na seems to sometimes be definite, and there are also examples where en seems to be indefinite or is used with a proper noun:
The most detailed description of en appears in a discussion of the origin of the name Roheryn (PE17/97), given here in full due to its importance to this topic. The note is undated, but I surmise that it may have been written in the mid-to-late-1960s given that one of the pages in the same bundle of notes discussed the root √RIK “try”, which itself probably dates to around 1966-7 (PE17/17, 93). However, Tolkien’s use of Hauð en ellas rather than Hauð en Elleth indicates that the note may instead be from 1950s before this name change (WJ/95 and WJ/148 note on §301).
This rendition of the quote is based on Christopher Gilson’s reading from PE17, including his editorial additions. It is hereafter referred to as Sindarin Genitive Notes (SGN):
Roheryn, softened Gondor form = pure S Roch(ch)eruin > Rocheryn = “Steed of the Lady” for {Roch ena Cheruin >>} Roch e Cheruin, Aragorn’s steed given him by Arwen.
Note 3 forms of Genitive in Old Sindarin. Without ending (in noun or article) follows an absolute noun and is then adjectival, implying any kind of relationship. So roch heryn = Lady(’s) horse, because connected with the lady (Arwen). Purely possessive (singular) is -a, plural -on.
And added after — old ī female, hence often in fem[inine] -uin.
Subject[ive] = ān/n, -āna. {roch heryna >> roch na cheryna >>} roch na-heryna, the horse of (the) Lady. Objective. dagnir an Glaurung, Slayer of Glaurung = he who slew Glaurung. an preceded article in form {ani >>} eni.
mellyn enin {Eleðrim >> Edhellon >>} Edhellion.
The possessive has article {en(n)a >>} ena usually, especially later, reduced to en before vowel (not when g is lost), na before consonant. Plural is enan > en n/, nan.
Hauð en ellas. nan ellas.
i·mbair en {Thor >>} N(d)engin, the houses of {the Dead >>} the Slain.
i·m(b)air en N(d)engin.
{glîm >>} glim {aiwon >> aewon >> aewion >>} maewion, maewia, (the) voices of gulls.
lais geledhion, or galaðon, the leaves of trees.-a is gen[itive] {ā >>} ō. {ion < ī + ām} ion is ia < g[enitive] iōm, later n restored.
X DON’T have inflected genitive! (PE17/97)
In the margin Tolkien had {Tor} >> Taur, apparently the unmutated form of Thor in i·mbair en Thor “the houses of the Dead”, which was rejected and replaced by i·m(b)air en N(d)engin “the houses of the Slain”.
In SGN Tolkien explored several distinct genitive constructions from (Old) Sindarin:
It is not clear how many of these genitive constructions were intended to co-exist. For example, Tolkien seems to have abandoned the possessive suffix -a/-ion. Of most interest to this discussion are definite genitive eni < ani and the possessive article ena, perhaps derived from adjectival *ina. Both would explain the singular genitive marker e(n) seen elsewhere in Tolkien’s writings, since the final vowel i or a would vanish at morpheme boundaries.
These ani and ena (< ina?) could be two competing etymological explanations for e(n) that Tolkien was considering. However, it is also possible that both these Old Sindarin constructs contributed to the modern use of e(n). Assuming all of these forms and functions co-existed, they have multiple (sometimes conflicting) uses:
Based on en {Thor >>} N(d)engin “of {the Dead >>} the Slain”, it seems the plural of en n/ could be used before consonants, so perhaps possessive article en(a) was the prevocalic form only in the singular. Given all these conflicting functions, it seems likely this system would regularize in modern Sindarin usage, but it is unclear exactly what this regularization would be.
There are numerous appearances of e(n) in Tolkien’s writings of the 1950s and 60s. To begin with, we pretty consistently see en before singular nouns beginning with a vowel. These could be either of the SGN genitive markers, pre-vocalic possessive article en(a) or definite genitive en(i) with vowel loss:
Several examples have ambiguous mutations, due to the presence of ancient nasalized stops mb, nd, ñg or an initial h:
The last two are also plural forms, increasing the ambiguity (see plurals below).
There is one clear example of a vocalic mutation:
There is one clear example of a nasal mutation, albeit in a rejected phrase:
Since this example was revised to a plural en N(d)engin, this deleted form may not be representative of the expected singular mutation.
A few examples seem to have non-mutation of voiced stops:
The last two are ambiguous, because they could be e-Ndanwedh or e-Nglad with orthographic variations. As noted previously, orthographic ambiguities make it hard to interpret Tolkien’s intent in some places.
There are a few examples of en before consonants that would not mutate, and all but one preserves the n:
The example en-Nibin-noeg is also plural (see below). The last example e·’Rach is especially curious, since the ’ seems to indicate some kind of mutation, but we don’t know if the unmutated form is *rach, *rhach or *grach.
Finally, there are three more plural forms without mutations:
The last two examples are from the Túrin Wrapper of the late 1940s or early 1950s, and perhaps en ir and en i were earlier iterations of plural enin from SGN, as given in the second example. For completeness, here are the three other plurals mentioned previously in this section.
There is enough variation in both the singular and plural examples of e(n) that there is a good chance some of them are not part of the same paradigm and thus not necessarily compatible with SGN.
A final set of interesting forms (all early variants of Haudh-en-Ndengin) appear in the Grey Annals from 1950-51:
The ina from the first example above may be an en form that has not undergone a-affection, and could hint at Tolkien’s earliest thoughts on the phonological origin of en. If so, it may connect back to emphatic/adjectival ina seen in documents on Quenya Personal Pronouns, as noted above (PE23/86, 105). Since it is plural and causes nasal mutation, the underlying form is probably ina(n).
The Grey Annals may also have the earliest appearance of en in the name Hauð-en-Ellas (WJ/92), later revised to Hauð-en-Elleth (WJ/95, 148, 267). As for Haudh-en-Ndengin, this form first appeared in Tale of the Children of Húrin (the “Narn” version as Christopher Tolkien called it) from later in the 1950s along with Haudh-en-Nirnaeth (WJ/169). Christopher Tolkien included both these forms in The Silmarillion as published, though apparently his father had Haudh-en-Nirnaeth as the “final” form in the Narn, so Haudh-en-Ndengin was omitted from the 2007 publication of The Children of Húrin.
Just like in the 1930s, there are numerous examples in the 1950s and 60s of i(n) being used in definite genitival expressions, almost always with plurals. These may be definite juxtapositional genitives or i(n) may be functioning as a distinct genitive marker:
The only singular examples are Athrad i-Nogoth which was revised to plural Athrad i-Negyth (WJ/338), and Condir i Drann (SD/129). Where glosses appear, they are almost always definite, with the exceptions of (a) inverted word order like “Dwarf-Road” and “Drúadan Forest” or (b) Tol-in-Gaurhoth “Isle of Werewolves”, which in Tolkien’s writings from the 1930s used na so the gloss could be a holdover of its previous indefinite form.
In most cases, it is impossible to tell if these examples are juxtapositional genitives that happen to have a definite article, or if the i(n) is a plural form of some other genitive marker like e(n). There is one place whether the latter is almost certainly true, in a list of names for Silmarillion tales (NST) probably from the 1950s (MR/373), discussed in more detail below in the section on mixed mutation.
In CEA itself, there are no genitival examples, since the main topic of that document is the definite article. However, since e(n) was the singular definite article in this document, it is unlikely to have served as a distinct genitive marker per se. Presumably definite genitives in this paradigm would have used e(n) [singular] or i(n) [plural] as a sort of definite juxtapositional genitive. Perhaps na would be retained for indefinite genitives, but we have no examples.
Summarizing the above information, I believe genitive articles went through the following conceptual stages:
1910s-20s: Gnomish had a genitive article na(n) [pre-consontal vs. pre-vocalic] with an indefinite prefixal variant a(n)-, both causing vocalic mutation. There was no distinct plural form at this stage. Attestations in the 1920s are rare enough that no conclusions can be drawn other beyond the fact that na(n) and a(n) were preserved in some form.
1930s-40s: Noldorin of the 1930s had a genitive sign na(n) [pre-consontal vs. pre-vocalic], still without a plural variant, having some abnormal (possibly mixed) mutations as discussed in the next section. At this stage na(n) seems to be indefinite. A prefixal variant an- was mentioned in The Etymologies. Drafts of The Lord of the Rings of the 1940s had a genitive sign a causing nasal mutation, which seems to be mostly definite. It may also be that i(n) functioned as a definite genitive marker in this period, possibly as a plural of na(n) or a(n), though genitival a also appeared before plural nouns.
1950s-60s: Sindarin retained genitival na(n), which in Quenya Notes (QN) from 1957 caused nasal mutation. In actual examples, definiteness and mutations seems to have varied in this period, but in a couple places there seems to be a definite plural variant nia, which in one example was revised to en. In one place there was a form ina, also later revised to en.
The form e(n) regularly appeared as a genitive marker in this period, inconsistently singular and plural, definite and indefinite, probably representing various paradigms. In Sindarin Genitive Notes (SGN) from the mid-1960s (or possibly earlier), Tolkien gave two (Old Sindarin) origins for en: one as a definite form en(i) < ani for an (archaic?) indefinite subjective genitive an, and another as a possessive article en(a) with no phonological origin provided but perhaps from *ina. In SGN both forms had distinct plurals enin and en n/ respectively. Attested examples of genitival e(n) from the 1950s and 60s are not always consistent with the SGN paradigm.
The form i(n) also regularly appeared in genitival constructions in this period, almost always definite and plural. For individual examples, it is hard to say whether these were definite juxtapositional genitives or if i(n) was a plural form of e(n), but there is at least one document (MR/373) where i(n) was almost certainly the plural of e(n), as discussed in the next section.
Finally in CEA from 1969, e(n) became the singular definite article and thus presumably no longer functioned specifically as a genitive marker, though likely it could still appear in definite juxtapositional genitives.
In his 2007 book Gateway to Sindarin, David Salo described a theory of mixed mutation for Sindarin en (GS/79-81). David Salo posited that Sindarin en was based on Old Sindarin ena < *ina, an astute guess since this was before the publication of PE17. Salo’s theory was that en induced vocalic mutation on following words up to the point where the vowel was lost at the morpheme boundary, after which the nasal n came into contact with the following word, inducing the effects of nasal mutation. He posited that in mixed mutation:
The last (apparent) non-mutation differs from ordinary vocalic mutation (d → dh) and ordinary nasal mutation (d → n) because both historical phonological processes were interrupted by shifts between vocalic mutational effects and nasal mutational effects. Using Salo’s example e·dant “of the fall” with references to Salo’s phonological rules from Chapter 4 of his book, the (somewhat simplified) phonological developments were as follows:
These simplified phonological developments ignore sound changes not directly related to mixed mutation. In the above, the voiced stop became a voiced spirant after the vowel (§4.86), but was re-stopped when it came in contact with the nasal (§4.125), and then the nasal vanished at the morpheme boundary (§4.187), resulting in an apparent non-mutation. This nasal-vanishing is analogous to the vanishing of nasals before mutated voiceless stops in nasal mutation, as seen in definite plurals like i·pheriain “the halflings”. It must therefore be limited to similar conditions like morpheme boundaries, since medial nasal-stop combinations became long (voiced or voiceless) nasals.
Salo’s theory of mixed mutation is compelling, but begs the question of what evidence there is that Tolkien imagined a similar mutational process, especially evidence that appeared after the publication of Salo’s book.
There is one place where Tolkien had a group of en-forms appearing together, in a set of various names for Silmarillion tales (MR/373). Presumably these names were part of an internally consistent paradigm, hereafter designated NST:
For these story names, Tolkien used e(n) before singular nouns and i(n) before plurals, with n only before vowels. Also of note is that in this paradigm, e(n) and i(n) appear before definite nouns but not proper names like Beren, Barahir, Húrin, Morgoth, and Gondolin. Thus e(n) and i(n) both seem to be markers of definiteness, and very likely also markers of the genitive. Assuming this reasoning is correct, in the NST paradigm i(n) is probably the plural of e(n), as opposed to enin and en n/ as seen in the SGN paradigm.
These story names are undated, but according to Christopher Tolkien they were “carefully typed but not on his later typewriter”. This, together with the fact that Tolkien uses ar for “and” rather than a(h) or a(ð), makes me think this document is from the 1950s.
In the NST paradigm we see the following mutations:
I believe the NST paradigm was David Salo’s main source for his theory of mixed mutation, especially given his use of e·dant as one of his main examples of the relevant sound changes and his theory that in was the plural of genitive article en (GS/149). Clearly Salo also looked at examples elsewhere, though he was limited to material published prior to 2007. For example, all discussions of alternate plural forms of en such as enin or en /n (PE17/97, VT50/5) were published after Gateway to Sindarin was written, so Salo can be forgiven for glossing over less obvious counter-examples like Haudh-en-Ndengin and Bar-en-Nibin-noeg.
As such, the NST paradigm is consistent with Salo’s theory of mixed mutation. Salo further hypothesized that the unmutated form of ’rach was rhach “curse” (GS/284), consistent with his notion that voiceless rh would be voiced in mixed mutation, but consonant clusters like gr would be preserved.
In the system described above, the most distinctive feature of mixed mutation for en is the absense of mutation for voiced stops. These can be used to identify possible examples and counter-examples to mixed mutation. Several further examples were known to David Salo:
Salo explained the first example as an orthographic variation of e-Ndanweth (GS/248) where the initial element was from primitive √NDAN “back” + ✶wedā “bond”, a plausible theory since these primitive forms appeared in The Etymologies (LR/375, 397). He explained the second example by assuming n was preserved before consonant clusters (GS/81). He explained the third example as another derivation from an ancient nasalized stop (GS/247), and daedelos was indeed confirmed as a derivative of √NDAY (PE17/151) after Salo finished his book.
An interesting pair of examples from around 1965 were also published after Salo’s book:
Here Tolkien seems to be vacillating between non-mutation of g versus nasal mutation to ng. An initial ng would be expected if Goloðrim was either (a) plural or (b) derived from √ÑGOL, both of which are known to be true from discussions elsewhere. However, perhaps in this moment Tolkien imagined that ancient nasalized stop ñg could mutate to simple g even for plural variants of en, so e·Goloðrim could be a transient example of non-mutation like e·dant.
Some examples that Salo was likely aware of but did not examine in his book were the apparent non-mutation of genitive na(n) from the 1930s:
Salo did not consider these names because they were all superseded by revised names in The Silmarillion as published: Taur-nu-Fuin, Tol-in-Gaurhoth, Haudh-en-Ndengin (S/155, 156, 197). However these apparent non-mutations from the 1930s could also be something like mixed mutation for the earlier genitive marker na(n) from the 1930s. The mechanism is not clear, but perhaps the actual primitive form was nan(a) rather than na(n). Non-mutating Dor-na-Daerachas “Land of Great Dread” from a 1971 addendum to The Lord of Rings maps could be similar (WJ/187).
Finally, Salo’s theory of mixed mutation hinges on the restopping of voiced spirants like ð after nasals at morpheme boundaries, to explain e·dant rather than e·ðant. If this was a feature of Tolkien’s Sindarin, there may also be evidence of it in compounds, the same way that ifant and Arathorn provide further evidence of nasal-loss before voiceless spirants (see above). One possible example is Baranduin “Brown River” = baran + duin (LotR/1138; VT48/23). If this underwent the normal medial developments, we would expect to see Barannuin. However, this is not a perfect analog of Salo’s mixed mutation, because we would expect **Baraduin under such a system.
One of the mutations previously mentioned in this essay but not fully explored are those of dative an in Quenya Notes (QN) from 1957, with the relevant portion of the quote repeated here:
S an, dative chiefly with pronouns or persons. < ana, hence vocalic mutation, {but preserves g, d} but takes form m before m, b. “to, for”. dative after word give ... S {ónen an >> ōn anim >>} ōn annin (PE17/147).
Here the base mutation seems to be vocalic, but with some exceptions. If dative an was derived from ana as indicated in QN, then it could have undergone phonological developments similar to genitive en derived from ena or eni, and thus could likewise have something like mixed mutation. If so, the deleted phrase “but preserves g, d” could indicate Tolkien was considering a system similar to Salo’s mixed mutation, since the preservation of voiced stops like g, d is the main difference between mixed and normal vocalic mutation.
Most attested Sindarin datives are combinations with oblique pronouns such as anim, annin, ammen, but there is one example that hints at possible non-mutation after an: gurth an Glamhoth “Death to the din-horde” (UT/39, 54). This example aligns with Salo’s system of mixed mutation, though Salo himself assumed dative an induced nasal mutation (GS/81, 140), as was the case in the King’s Letter: a Pherhael “to Samwise” (SD/129). Unfortunately, orthographic ambiguity means that gurth an Glamhoth could also represent a nasal mutation: gurth a-Nglamhoth. The variations may also be because Glamhoth is plural rather than singular.
Nevertheless I think the material published after Salo’s book further supports the possibility that at various points in time Tolkien had something like Salo’s mixed mutation for genitive e(n) and na(n) [the latter especially in the 1930s]. It may also have applied to dative an as described in QN, though not for all uses of the dative given the nasal mutations in the King’s Letter. We don’t have enough information to know if Tolkien’s system agreed in all details with Salo’s, or the conditions in which mixed mutation applied.
As an aside, after the deletion of “{but preserves g, d}” Tolkien’s revised mutation for an reads: “hence vocalic mutation, but takes form m before m, b” (PE17/147). This is remarkably similar to the mutation system of the emphatic article as described in The Anaphoric Particle, rough notes that are either a draft or a follow up to CEA:
CE inā ... was used gen[erally] in S. inā > en(a) > en, and a pl. in(i) > in, both with vocalic mutation ... this before cons[onants] would have produced S en followed by vocalic mut[ation] and assimilation of n. So *en(a) parth “the garden” > *em-barth (PE23/141).
So even if Tolkien had something like mixed mutation for na and en in the 1940s and 50s, it seems likely that at some point he abandoned it in favor of ordinary vocalic mutation (with some nasal survival and assimilations).
As a final point for this section, it is worth considering how mixed mutation for the genitive marker e(n) compares to the 1969 CEA paradigm where e(n) was a singular definite article. We can use the NST paradigm (MR/373) as a baseline:
If we assume the above are definite articles as they were in CEA, the only clear disagreement with the CEA system is e·Dant “*the Fall” rather than e·Dhant, especially if we ignore e·’Rach whose unmutated form is unclear. If mixed mutation for e(n) was the norm prior to CEA, then the revisions of CEA would have only minimal impact, since the only mutation undergoing significant changes would be those for voiced stops.
Another possible disagreement is e·mbar “*the house” < √MBAR. In the NST paradigm, we don’t know if mb represents [mb] or is an orthographic represent of [m], but I suspect the former. In Tolkien’s earlier writings, he seems to vacillate between mb (1910s: PE11/9), m (1920s: PE13/120), and mm (1930s: PE19/20) as the vocalic mutation of words with ancient nasalized stop mb. The long nasal mm in particular only makes sense as a phonetic rather than orthographic representation. However, by deciding in CEA that mb was sometime used as an orthographic representation of the actual mutation m, Tolkien hedged his bets by providing an alternate explanation for examples where mb appeared in writing.
Therefore, I believe that Tolkien intentionally crafted his system of mutations in CEA so that e(n) as a singular definite article was mostly backwards compatible with e(n) as a singular genitive article, even though it may originally have used mixed mutation rather than vocalic mutation.
In many respects, the system presented in Common Eldarin Article (CEA) is compatible with Tolkien’s Sindarin ideas from earlier writings. Most of the vocalic and nasal mutations are consistent with what we know about Sindarin of the 1950s and 60s, and in many cases earlier iterations of the language as well. It also confirms some items that were suspected but not well documented, such as voiceless lh, rh archaically mutating to thl, thr, but remaining unmutated in modern Sindarin.
However, CEA departs from Tolkien’s previous system of Sindarin articles and mutations in several obvious ways:
1) The singular definite article in CEA is e(n), as opposed to previous i. This likely also implies the removal of genitive article e(n).
2) The full plural article in survives before mutated voiceless stops (in·pherth), as opposed to previously where the n was lost (i·pheriain).
Less obvious changes are:
3) Labial nasal m does not always undergo vocalic mutation in modern Sindarin (PE23/138 note #12). Given examples of non-mutation in The Lord of the Rings, this may have been the rule earlier, or Tolkien may have contrived this new rule to explain inconsistencies in The Lord of the Rings.
4) Ancient nasalized clusters reduce to simple nasals in nasal mutation (PE23/139). Something similar may happen for vocalic mutation, but the documentation in CEA is unclear. Tolkien seems to have introduced an orthographic convention where simple nasals m, n, ñ are sometimes represented in writing as mb, nd, ñg to better distinguish them from original simple nasals m, n. This also explains cases where such clusters appeared in Tolkien’s earlier writings such as i mbas (VT44/21), di’nguruthos (LotR/729), and Taur e-Ndaedelos (LotR/1134). While it is possible these earlier examples were also orthographic representations of simple nasals, I think it is more likely that (at least part of the time) they represented preserved clusters in Tolkien’s earlier systems of mutation.
Most subtle of all are the changes to the phonological history of Sindarin underpinning his new system of mutations:
5) Nasals only vanish before voiceless spirants early in Sindarin’s history, limited to those spirants resulting from ancient aspirates. The same does not apply to voiceless spirants resulting from combinations of nasals and voiceless stops, which explains why the in survives in such mutations (PE23/137).
6) Combinations of nasals with voiceless spirants do not further evolve into voiceless nasals, which implies that Tolkien removed this sound change from Sindarin’s phonological history. Supporting evidence can be seen in contemporary inflected past tense forms which likewise retain combinations of nasals with voiceless spirants: tancher (PE23/138 note #12) and echanthel (VT47/38). This removal of voiceless nasals as a stage in Sindarin’s phonological developments is surprising but not entirely unprecedented: Gnomish of the 1910s had no such sound change, and there are hints Tolkien was considering a similar removal in Quenya Notes from 1957: “nt &c. no longer are held to yield nn” (PE17/185).
These modifications of Sindarin’s phonological history seem to be part of a general desire to simplify Sindarin’s system of sound changes, bringing mutations of initial consonants into closer alignment with medial sound changes. Tolkien’s previous system of nasal mutations and mixed mutations (assuming Tolkien had such mutations) both required specialized nasal losses only occurring at the beginning of words or morpheme boundaries. In the CEA system, the medial developments and the developments producing mutations are more similar to each other.
Where Tolkien did make changes to his system of articles and mutations, they seem to have been deliberately crafted to be (mostly) backwards compatible with his earlier ideas. For example, Sindarin names of Silmarillion tales from MR/373 very likely used genitive articles: singular e(n) and plural i(n). However, they are almost entirely compatible with the CEA system if the genitive articles are reinterpreted as definite articles in juxtapositional genitives, the only big exception being e·dant “the fall” rather than e·dhant.
Despite this, there are a couple points where the CEA system is incompatible with The Lord of the Rings. In Appendix A, i is used as a singular article in i-Estel (LotR/1061) rather an en-Estel as it would be in the CEA system. There is a vague statement in CEA where Tolkien said “[singular] i used orig[inally] only before e < ge” (PE23/135). This “orig[inally]” might be interpreted as meaning there could be other situations in modern Sindarin where singular i could be used before other words beginning with e. However, in the related document “The anaphoric particle” Tolkien gave as examples i-êl “the weeping” (presumably from *gêl) versus en-êl “the star [êl]” (PE23/141). Therefore, it seems likelier to me that Tolkien intended words beginning with e to also use en as their definite article in the CEA system.
Even more notable are the differences in nasal-loss before mutated voiceless stops. The Lord of the Rings had i thiw (LotR/305), i Pheriannath (LotR/768), and i Pheriain (LotR/953). Note that towards the beginning of CEA Tolkien has a chart indicating that the pre-consonantal form of the definite article was i (PE23/135), and towards the beginning of the discussion of nasal mutations Tolkien again says “In the plural (a) i was followed by nasal mutation” (PE23/137). But when he gets into the particular mutations of specific consonants, he breaks this rule, not only for voiceless stops but also in before r where the result is iðr (PE23/137). Since Tolkien devotes two full paragraphs of text explaining why nasals survived before mutated voiceless stops (PE23/137), I think it is likely he intended this to be an exception to the general rule. This is especially likely since he called out other cases where modern Sindarin mutations further regularized, such as rh, lh → archaic i·thr, i·thl versus modern i·rh, i·lh, and gw → archaic in·w versus modern i·ñw.
Superficially it may seem surprising that Tolkien missed such obvious counter-examples to his new system of articles and mutations, especially since he made several references to examples from The Lord of the Rings and it was clear he was comparing his new system to the books. However, this would not be the first time he invented a new system that contradicted his published works, or even deliberately changed previously published forms, most famously omentielmo becoming omentielvo between the 1st and 2nd edition of The Lord of the Rings.
There are even examples of such missed references in notes associated with CEA itself. As part of his consideration for abandoning the vocalic mutation of m, Tolkien wrote:
Great relief if mutation of m could be avoided. It actually occurs v[ery] rarely in LR. Only case I can find is Menelvagor, I 91; silivren [<] silmarin, I 250; but not used in Imloth Melui, III 142, 244! (PE23/138 note #12)
Here Tolkien completely misses one of the most notable examples of the mutation of m, namely na vedui “at last [medui]” from Glorfindel’s greeting to Aragorn (LotR/209). These may seem like “obvious” mistakes to us, but we live in an era of searchable digital texts and a community of peer-reviewed Tolkien scholarship, while Tolkien was working off paper documents and his own memory.
Caution must be taken before overgeneralizing from CEA itself, however. The document ends with a title for an unwritten section “Derivation of the forms of the S. definite article”. Thus CEA itself is unfinished, since it appears Tolkien was ready to embark on further exploration of the historical derivation of the articles. It is possible that Tolkien would have reversed himself on some of the ideas appearing in CEA after digging deeper into their phonological ramifications. Unfortunately, as is often the case with Tolkien, we have no way of being certain.