On Telerin

Ellapesta

J.R.R. Tolkien frequently used the term “Elf-latin” or “Elven-latin” in his works, with various meanings. Most often he was comparing the role of an Elvish language in Middle-earth to the cultural role of Latin, particularly as it existed as a language of learning and liturgy after the fall of the Roman Empire, in medieval and early modern Europe:

  • “The High-Elven…. had become, as it were, an ‘Elven-latin’, still used for ceremony, and for high matters of lore and song…” (The Lord of the Rings, Appendix F);
  • “Q[uenya] was known to the learned in Gondor at least as well as Latin still is in W[estern] Europe.” (The Letters of J.R.R. Tolkien p. 425, #347);
  • “the nearest equivalents to Latin and Greek in Shire-lore were the Elvish tongues”; (The Lord of the Rings, Appendix F).

He might also refer to the tone and feeling of the language, related but not necessarily identical to its use:

  • “Avallonian is… more august, more ancient, and, well, sacred and liturgical. I used to call it the Elven-Latin.” (The Notion Club Papers, Sauron Defeated p. 241)

He also made the comparison in terms of orthography:

  • “The High-elven Quenya has been spelt as much like Latin as its sounds allowed.” (The Lord of the Rings, Appendix E)
  • “The archaic language of lore is meant to be a kind of ‘Elven-latin’, and by transcribing it into a spelling closely resembling that of Latin… the similarity to Latin has been increased ocularly.” (Letters, p. 176 #144)

And in terms of overall phonological relationships:

  • “Sindarin…is in fact constructed deliberately to…. have a relation to High-elven similar to that existing between British… and Latin.” (Letters p. 219, #165.)

Tolkien had a strong background in the classics, and learned Latin and Greek as a child at King Edward’s School (Letters, p. 213, #163) where “the chief contacts with poetry were when one was made to try and translate it into Latin.” Tolkien was later strongly influenced by Gothic and Finnish; but, as he wrote in 1955, “Linguistic taste changes…. or oscillates between poles. Latin and the British type of Celtic have it now…” (Letters, p. 214, #163).

Sindarin was of course modelled after “the British type of Celtic.” But what of Latin?

Tolkien explicitly said “Quenya is the Elf-latin” (Lammasethen). But any careful scrutiny of Quenya reveals that it does not much resemble Latin, certainly not phonologically (nor, indeed, morphologically or syntactically). Tolkien wrote, “The ingredients in Quenya are various, but worked out in a self-consistent character not precisely like any language that I know.” (Parma Eldalamberon #17, p. 135). When Tolkien noted that Quenya was spelt like Latin “as much… as its sounds allowed,” he was admitting that there were strong differences in the sound-systems of the two languages.

Goldōrin ar Rōmārin

In what follows, * precedes an historical reconstruction, ** precedes a counterfactual one, and ! precedes a form that has plausible arguments for its existence, but is not actually recorded.

Latin, considered in the context of Indo-European languages, had (at the time of the late Republic and early Empire) a remarkably simple phonological system. It possessed five vowels (a, e, i, o, u) with short and long variants; a small handful of diphthongs, of which only ae, au, and the relatively rare oe continued Indo-European diphthongs (a few others, such as eu, oi, ui arose from later coalescence of originally separate vowels, e.g. neuter < ne-uter, coitio < co-itio, cui < *kʷoiei); and the 15 consonants b, k (spelt c, or q when before a w-sound), d, f, g, h, j (spelt i), l, m, n, p, r, s, t, w (now conventionally spelt v, but never in classical times distinct in spelling from the vowel u, both being spelt with the written letter V).

Quenya, by contrast, though it has the same five vowels with length distinctions (which variants are not, however, used to distinguish roots as in Latin) has a richer supply of basic diphthongs (iu, eu, au, ai, oi, ui, all in common use) and a much larger number of consonants. Counting the number of consonants in Quenya is difficult, since it is not always obvious which sounds are distinct phonemes, which are underlyingly sequences of two sequential sounds, and which are allophones with a restricted range; but it seems reasonable to state that there are about 22:
labials p, f, m, b (the last only after m or l), v;
labio-velars w (later partly merged with v), hw (i.e. [ʍ]);
velars k, ŋ, g (only after ŋ);
palatals y, hy (i.e. [ç];
coronals t, θ (later merged with s), n, d (only after n, r, l), r, hr (i.e. [r ̥]), l, hl (i.e. [l ̥]), s;
glottal h (or velar [x]).
[z] was never a separate phoneme, being first an allophone of s or d, then merging with r.

Quenya also makes considerable use of the labialized velars kʷ, ŋʷ, and (only after ŋ) and the palatalized coronals ty, ny, and ly, all of which except could occur initially. Other initial consonant combinations are not found. Ps-, ks- are asserted to occur as initial clusters (as in Greek); but most Quenya words of this type show ips-, iks- (ipsin “fine thread” < *psin < *spin-, ixal “shadow” < *ksal < *skal- Parma Eldalamberon #17, pp. 17, 184 ; a verbal root psar- appears in PE #18, p. 94, but no derived forms are given).

The status of [d], [g], [b], and [gʷ] is the most problematic, due to their extremely restricted range; it is reasonable to consider [b] and [v] allophones of a single phoneme (which could equally well be posited as /b/ or /v/), but [d] has to be given phonemic status, and, despite their narrow distribution, there seem to be insuperable obstacles to regarding [g] and [gʷ] as anything other than the surface realizations of underlying /g/ and /gw/.

The differences from Latin, especially with respect to the consonants, are striking. Latin lacks palatalized consonants altogether, and even the sequences *tj, *kj, *dj, *nj, *lj, *sj, etc. do not occur — where they had occurred in Proto-Italic, in Latin they had become -ti-, -ci-, -di-, -ni-, -li-, -ri- (e.g., in verbs like patior, facio, gradior, venio, salio, orior); not until a very late stage in the development of Latin did the vowel i (unstressed) between a consonant and a vowel change back to [j], which subsequently produced a variety of palatalized consonants in the Romance languages.

In Latin, unlike Quenya, the sounds of b, d, g are common, and frequently occur between vowels (e.g. labor, fides, vigor) as well as initially (bos, dens, gens).

Latin consonant clusters are common, especially those of stops and f + r or l, or s + stop. Quenya avoids all initial consonant clusters except those palatal and labio-velar sounds which can be interpreted as underlyingly /tj/, /kw/, etc., and the rare and evanescent clusters ps-, ks-.

Latin, again unlike Quenya, lacked a distinction between /w/ and /v/. In the Latin of the later Roman Empire, /w/ began to show a phonetic differentiation, becoming [v] initially and in most medial positions, but remaining in the sequences [kʷ], [gʷ], [sw].

Latin had fewer labio-velars than Quenya; practically they were limited to [kʷ], [gʷ], [w] (before the change of the last to [v]), and in native Latin words [gʷ] was practically limited to the sequence [ŋgʷ] in words like anguis, inguen, lingua. But in the latest stages of Latin, after [w] had come to be pronounced [v], Latin-speakers encountered a new [w] in the Germanic languages of the Goths, Franks, Lombards, and so forth. This [w], in words borrowed into Late Latin, was sometimes represented by the written letters GV (i.e. [gʷ]), as well as the innovative, though confusing, written letters VV (i.e. [w]). Latin did not have [ʍ] (Quenya hw).

Latin, unlike Quenya never had initial velar or labio-velar nasals; the velar nasal [ŋ] was restricted to internal positions before [k], [kʷ], [g], and [gʷ].

Quenya final consonants are limited to coronals: t, l, s, n, r. These were also favored in Latin, but other sounds occurred, some fairly rarely or in restricted word-classes (e.g., b, c, d); on the other hand, final m was very frequent. Latin also had multiple consonants at the end of a word — usually a consonant + s ([ks] (spelt x), ps, bs, ns, rs, ls), and also the 3rd person plural active desinence -nt; whereas Quenya is only allowed one consonant word-finally, except in archaic -nt (e.g. in the ‘dative’ dual termination.)

Latin tolerated, and indeed very frequently has, the sequence [kt], especially in past participles like dictus, factus, lectus, and many words derived from the same participial stem (e.g. from fact-: factum, factor, factio, facto, factito). In Quenya [kt] is forbidden, and ht (pronounced [xt] or [çt]) appears instead.

It is thus obvious that — Tolkien’s minor efforts at ‘Latinizing’ Quenya orthography aside (essentially limited to using c for [k] and qu for [kʷ]) there is no considerable phonological similarity between Latin and Quenya. As Tolkien wrote, Quenya is “not precisely like any language that I know,” and more clearly than any other of his constructed languages, reflects Tolkien’s phonetic predilections.

All or nearly all of Tolkien’s other constructed languages, however, are phonological imitations of some other real-world language. Khuzdul and Adûnaic resemble ancient Semitic languages (particularly those with simplified sound-systems, like Babylonian and Phœnician). The Danian of The Etymologies was clearly modelled on Old English; Taliska was modelled on Gothic. Sindarin was inspired, as Tolkien repeatedly noted, by the historical evolution of Welsh. In many points of detail, of course, it stands apart, and is rather unlike Welsh in its treatment of the vowels; yet at the same time it displays an overall character which remains very close to Welsh, and reflects Tolkien’s ‘linguistic taste’.

What, then, happened to Tolkien’s ‘linguistic taste’ for Latin, which gripped him in 1955, and presumably (due to ‘oscillation’) at other periods before and after?

The answer was somewhat of a surprise to me, though the evidence for it was (just barely) there.

Orperie

In the early 1950s, the family tree of Tolkien’s languages went through an overwhelming transformation which I call the Great Upheaval. The changes were complex, far-reaching, and in some ways destructive.

Prior to the Great Upheaval, ‘Quenya’ was the language of the Vanyar (formerly called the Lindar); ‘Old Noldorin’ was the language spoken by the Noldor in Aman; ‘Exilic Noldorin’ — a language almost identical in phonetic character and many other respects to the later Sindarin — was the language spoken by the Noldor in Beleriand; ‘Doriathrin Ilkorin’ was the language of Thingol of Doriath; and ‘Danian’ was the language of the elves of Ossiriand.

After the Great Upheaval, Quenya was the language of the Noldor (with the Vanyar speaking a closely related dialect, differing in details of pronunciation), in both Beleriand and Aman; ‘Old Noldorin’ became an ancestral stage of the Sindarin language spoken in Beleriand before the return of the Noldor; ‘Exilic Noldorin’ became Sindarin proper. ‘Ilkorin’ and ‘Danian’ seem to have disappeared; the very scanty information on the Nandorin tongues (which took the place of ‘Danian’) suggests a language of a quite different phonetic character from either.

Also vanished were three intimated Avarin (formerly ‘Lemberin’) tongues spoken in Middle-earth; these were originally intended to have a resemblance in sound to Goidelic, Baltic, and Finnic languages, but the list of Avarin ethnonyms in the essay Quendi and Eldar shows an entirely different set of developments.

The new Nandorin and Avarin languages, insofar as publications to this point have provided evidence, were never elaborated to the extent of the pre-Great Upheaval languages. Instead, Tolkien’s essays and notes of the 1950s and 60s focus on only three Eldarin languages: Quenya, Sindarin, and Telerin.

Ortāle Lindarinua

Prior to the Great Upheaval, Telerin had been a not-terribly-interesting side branch of the Eldarin language family. The Teleri themselves served little purpose in the history of the Elves other than to lag behind on the march from Cuiviénen, and later to be slaughtered by the Feanorian Noldor at Alqualondë. Their language was supposed to be more closely related to ‘Ilkorin’ than to Quenya or Noldorin, though there is little phonetic evidence in support of that notion. The evidence is, however, very scant. In The Etymologies, 25 Telerin words and names are found, considerable less than what is recorded for Ilkorin, and barely more than for the evanescent Danian language; this suggests its relative unimportance at the time.

After the Great Upheaval, Telerin suddenly achieved prominence in Tolkien’s linguistic writings. Mostly this had to do with its newfound importance in the family tree: it was now more closely related to Sindarin than to Quenya, yet equally as archaic as Quenya, and so on many technical points provided a tertium quid between those two languages, which could shed light on developments in both.

But it also offered Tolkien an opportunity to indulge one of his ‘linguistic tastes’. For Telerin, as it turns out, is phonologically based on Latin.

The evidence for this is very direct. It might have been deduced based on the evidence of the word-data alone, but only rather ambiguously; and the prevalence of Latinisms in European languages is so pervasive that a specifically Latin phonological style tends to be invisible — it is, after all, exactly what one would expect. The Latin basis of Telerin is, however, now beyond doubt.

In an early (pre-Great Upheaval) text, probably of the late 1930s (and so slightly prior to The Etymologies) published under the title of Comparative Tables in Parma Eldalamberon #19, Tolkien describes Telerin as “of an approximately Latin type, but with labialization of qu > p.” The main part of this text is a list of the primary developments of the sounds of “Primitive Quendian” in a series of languages: Quenya, Vanyarin (given as ‘Lindarin’), Telerin, Old Noldorin, Noldorin, Ilkorin, Danian, East Danian (Tolkien does not here use the term ‘Leikvian’, but the language meant is the same), Taliska, and West, North, and East Avarin. There is a great deal to be said about the rest of these languages, but in this essay I will focus on Telerin.

Lindārin ar Rōmārin

In this version of Telerin — and it must be remembered that this is a very early version, different in important aspects from later Telerin — the Primitive Quendian sound-system is greatly changed. Among the consonants palatalized velars (corresponding to Quenya’s palatalized coronals) are wholly merged with the plain velars; likewise, the labialized velars merge with the plain labials. Telerin thus turns Primitive Quendian’s five places of articulation into only three. The resulting system of consonant sounds is quite simple and recognizable: labials p, b, f, m, v; velars k (spelt c), g; palatal j (so written — most likely meaning [j], i.e. the sound of y in “yes”) coronals t, d, n, r, l, s; glottal h. These are exactly the 15 consonants found in Latin.

The use of the written letter c for [k] is interesting, as it foreshadows Tolkien’s eventual decision to substitute c for k when presenting Quenya words in The Lord of the Rings — although his previous usage, and his private usage in unpublished materials, was almost always to write Quenya [k] with the written letter k, in which case they often contrasted with Telerin words written in a similar manner, but with c for [k].

The letter v is used as a reflex of Primitive Quendian *w. It appears from later evidence that the letter is merely an orthographic variant, intended to be pronounced [w], inspired by the current use of v to represent Latin [w]. The evidence is the explicit statement, regarding the contrasting developments of Quenya *áwa > *áua > oa, and Telerin *awa > ava, that “v remains w in sound” (The War of the Jewels, p. 367) and the representation of Telerin v by w, not by v, in a transcription which is explicitly “spelt as Quenya” (Vinyar Tengwar #48, p. 6)

It is not, however, impossible that at some point it was intended to represents an actual sound change to [v]. It would be a plausible change, because the complete elimination of all labialized velars would leave [w] as a lone labio-velar with no closely related sounds; and because the presence of [f] as a voiceless labial (or labiodental) fricative would attract a similar sound to fill in the gap where a voiced counterpart to [f] would go, as indeed happened historically in Late Latin.

Other historical developments show a close resemblance to Latin. Primitive Quendian *tʰ (aspirated t) has the Telerin reflex (in this early version) f. Proto-Indo-European (as normally reconstructed) had no *tʰ, but it possessed a breathy-voiced *dʱ, which became *tʰ in Proto-Italic (and Proto-Greek), and in Italic first became a fricative *θ (like th in English ‘thorn’) and later, when initial, became f. For example, Proto-Indo-European *dʱuHmos “smoke” yields Sanskrit dhūmas, Greek thūmos (“spirit”), and Latin fumus ([fūmus]) – whence English ‘fume’. The intermediate form was *θūmos.

There is actually an exemplar of this particular change in The Etymologies — the Quenya name of King Thingol, was, according to that text, “Sindo” — reflecting older *[θindo] — but “in Telerian form Findo” (The Lost Road (LR) p. 392, root entry THIN-), showing a change in this stage of Telerin from *θ > f.

The same *tʰ is treated differently in the middle of a word; according to Comparative Tables, it becomes either d or b. This change looks mysterious, unless seen in the light of Telerin’s resemblance to Latin.

In Proto-Italic, Indo-European *bʱ and *dʱ followed partly converging courses. They first became *pʰ and *tʰ respectively; then, in most Italic languages, they merged as [f]. In Latin, however, the original distinction is still partly to be seen. Initially, as noted above, *tʰ became [f] and merged with [f] < *pʰ. Medially *pʰ became, first [f] and then voiced to (presumably) [v] which was subsequently stopped to [b] — a change partially parallel to the voicing of medial *s to [z] (which later became [r]). *tʰ also became [θ] > [f] > [v] > [b], but only under restricted conditions, namely when before or after the liquids l and r, or next to u. Although the effect is the same, the causes are likely different; probably [θl],[θr],[rθ] > [fl], [fr], [rf] as a form of dissimilation, so that two coronal sounds did not stand adjacent; whereas [uθ] > [uf] is probably a kind of assimilation, with the labial quality of the [u] carrying over to a following consonant. Examples are *rudʱros > *rutʰros > *ruθros > *rufros > *ruvros > *rubros > *ruberos > ruber “red”; *werdʱom > *wertʰom > *werθom > *werfom > *wervom > *werbom > werbum (verbum) “word.”

When not subject to the conditions above, [θ] evidently remained [θ], then when other medial fricatives voiced, it became [ð], which was then stopped to [d]: *bʱeidʱō > *pʰeitʰō > *feiθō > *feiðō > *feidō > *fīdō (fido) “I trust.”

There is an example of precisely this kind of change in The Etymologies, though its nature was not immediately evident before. In that text, the Telerin word for the beech tree is said to be bredele; but being derived from a root BERÉTH (LR:352), it must be from *berétʰ- with syncope of the first vowel > *bretʰ-, with the addition of a suffix -ele; then *bretʰele > *breθele > *breðele > bredele. Exilic Noldorin brethil, in the same entry, probably shows a form of the same suffix; in the root entry NEL- (LR:376) the word for beech is said to be brethel, with brethil as its plural. In later Sindarin brethil is apparently the birch tree, not the beech, though Tolkien evinced some confusion on this point; the name Fimbrethil (of an Entwife) being translated both “slender-beech” and “slender-birch.”

Also like Latin is the treatment of [j] following a consonant. As mentioned and exemplified above, internally Indo-European [j] became syllabic, turning into the vowel [i] after a consonant; also cf. *medʱjos > *metʰjos > *meθjos > *meðjos > *medjos > *medios > medius “middle.” Comparative Tables shows the internal reflexes of *kj, *kʰj, *gj, *ŋj to be -ci-; -gi- or -di-; -gi- or -di-; and -ŋgi- or -ndi-.

The first of these is comparable to Proto-Italic *-kj- > Latin -ci-, e.g. *θakjō > *fakjō > fakiō (facio) “I make.” The remainder are not the same as Latin. Latin *-gj- became not -gi- but a geminate [jj] (written, however, simply i) : e.g. *magjos > *majjos > majjus (maius) “greater.” *-ŋj- did not occur in Latin. *-gʱj- > *-kʰj- probably also became -j-, a possible example being *wegʱjā > *wekʰjā > *wexjā > *wehjā > *wejā > *weiā > *wīā > *wia (via) “road.”

A notable divergence between the Telerin of “Comparative Tables” and Latin comes in the treatment of *kʰ (and likewise *kʰj). Indo-European *gʱ, which became Proto-Italic *kʰ, later *x, did not share the medial voicing of *f, *θ, *s and the stopping of *v, *ð > b, d. Rather, *x became h, which — being not strictly a fricative but rather a lack of voice — did not share in the voicing of other fricatives. At least, that is one way to interpret the evidence: e.g. *wegʱō > *wekʰō > *wexō > *wehō (veho) “I carry, I transport.” The change to g *does* occur in Latin following a nasal, in which position it seems that the weakening of *x > h did not occur: *dʱingʱō > *tʰinkʰō > *θinxō > *θinɣō > *θingō > fingō (fingo) “I form, I fashion.”

In the “Comparative Tables” Telerin, *x apparently did voice to *ɣ, because the reflex of medial *-kʰ- shows up as -g-; this also explains *kʰj > -gi-. No examples of this particular change are found in published texts. It would presumably have shown up in words from roots ending in , like √mbakʰ- ‘exchange’ (LR:372); e.g. *mbakʰā “ware” (article for trade) would have become !baga.

Syllabic nasals were relatively common in Proto-Indo-European, as zero-grade variants of sequences such as en, em, on, om. They are much less common in Primitive Quendian, but they appear initially as enhancements of initial clusters of nasal+vowel (mb, nd, ŋg, ŋgʲ, ŋgʷ) which are quite common (and have no Indo-European counterpart). When not ‘enhanced’, they appear in Quenya as pure nasals (m, n, ŋ, nʲ, ŋʷ) and in Noldorin/Sindarin as stops (b, d, g). With syllabification of the nasal element, we find various sequences of vowel + nasal + stop: in Quenya umb-, and-, iŋg-, in(d)y-, uŋgʷ-; in Noldorin/Sindarin am(m)-, ann-, aŋg-. For example *m̩bart- produces Quenya umbar (genitive singular umbarto), Noldorin/Sindarin am(m)arth.

In Telerin the normal reflex (and the only one attested anywhere) is e+nasal-: emb-, end-, eŋg- . Comparative Tables shows the reflex of *ŋ̩gj- to be aŋg-, not eŋg-, which is exceptional, problematic, and in all likelihood due to confusion with the neighboring column, which gives the Old Noldorin reflexes, all with a+nasal. The Etymologies gives as an example Telerin endo < *ŋ̩gyō, cognate to Quenya indyo, Old Noldorin ango. In Latin as well, the reflexes of syllabic *n̩ and *m̩ were en and em; for example *tn̩tos “stretched” became *tentos > tentus; likewise *km̩tom became by assimilation *kn̩tom > *kentom > kentum (centum) “hundred,” and *dekm̩ became decem “ten.”

Other changes to consonant combinations in “Comparative Tables” also correspond to those of Latin; e.g. initial *dl > l, as in Latin: *dlongʱos > *dlonkʰos > *dlonxos > *dlonɣos > *dlongos > *longos > longus “long,” *dlakt > *lakt > lak (lac) “milk”; and initial *sr > fr, as in *srīgos > *θrīgos > frīgos > frīgus (frigus) “chill, coldness.”

In the vowel system, the resemblance to Latin (of a more archaic kind) is just as notable. According to Comparative Tables, Telerin ai at a later date became ae, and eventually a variant of e; this exact change took place in Latin, e.g. Old Latin *saitlom, Classical Latin saeculum, Late Latin sec(u)lum “age” (cf. English ‘secular’). Quendian *oi is shown merging with Telerin ui, and subsequently becoming long ū — a similar change occurred in Latin, e.g. *oinos > *ūnos > ūnus (unus) “one.” Telerin au is shown monophthongizing at a late date to a low back o-type vowel; this actually occurred twice in the history of Latin, once during the era of the Roman Republic, as a deprecated rusticism (thus one branch of the famous Claudian family pronounced and even spelled its name Clodius); and later in Proto-Romance, whereby Latin words like causa, aurum became Italian cosa, oro, French chose, or (“thing,” “gold”).

Tembesta Lindarinua

A small amount of additional information about Telerin is available in two texts close in time to Comparative Tables: the first version of the Tengwesta Qenderinwa (TQ1), published in Parma Eldalamberon #18, pp. 23-68, and the Outline of Phonetic Development (OPD), in Parma Eldalamberon #19, pp. 29-67. These two texts provide the Telerin words pentro ‘reciter, minstrel’, tancula ‘clasp, brooch’ (TQ1 pp.50-1), lauda ‘gluttonous eating’, hāda ‘piled mound’, Erde ‘the Valie Este’ (OPD p. 45), burbo ‘large fly’ (OPD p. 48).

These show the following phonological developments, compared to Quenya, with notes on Latin and other parallels below:

1) *kʷ > p (Q. qu): T. pentro, Q. quentaro

2) *-ntr- = -ntr- (Q. -ntar-, -ns-): T. pentro, Q. quentaro, quenso (OPD p. 40)

3) *l > ul /C_V (Q. al): T. tancula, Q. tankala

4) *-sd- > *-zd- > -rd- (Q. -st-): T. erde, Q. este

5) *-zb- > -rb- (Q. -ps-) : T. burbo, Q. pupso

6) *-bd- > -wd- (Q. -ps-): T. lauda, Q. lapsa

7) *-gd- > -ːd- (Q. -ht): T. hāda, Q. hahta

(1) Latin differed from Telerin by preserving the sounds [kʷ] and [gʷ], whereas Telerin converts them into labials. However, some Italic languages related to Latin, such as Oscan and Umbrian, had the same conversion of labialized velars to labials as Telerin, e.g. Umbrian panta for Latin quanta “how many.” The conversion to labials also took place in the variety of Late Latin ancestral to Romanian (e.g. lingua > Romanian limba “language”), though only partially and in a restricted environment.

(2) Latin preserved the cluster -ntr- when not final, e.g. intra, contra; but linter “boat” (?< *lintrs) genitive sg. lintris. T. pentro is < *kʷentrō.

(3) shows the addition of a syllable by the addition of a svarabhakti vowel -u- in a cluster of obstruent + l: *tanklā > tancula. This latter change is like that of Latin *pōtlom > *pōklom > pōkolom (pocolom) > pōkulum (poculum) “cup” (from a root √- “drink”). The word tancula was again used as an example in the later Tengwesta Quenderinwa, (Parma Eldalamberon #18, p. 100), written several years later.

(4) In Latin, the sequence -sd- did occur word internally. The result was the deletion of the -s- with compensatory lengthening of the preceding vowel; e.g. *nisdos (from *ni-sed-os, ‘down-seat, what one sits down on’) > *nizdos > *nīdos > nīdus (nidus) “nest.” The Telerin change *-sd- > *-zd- > -rd- (*ezdē > erde) is obviously different, though it does show a change of s to r which did occur (in other environments) in Latin. The change *-zb- > -rb- (*buzbō > burbo) is similar.

(5-7) exemplify exceptionally rare or difficult consonant combinations. The Telerin developments *labdā > lauda and *kʰagdā > hāda are sui generis; they both show a weakening of the voiced stops b, g, the first to a labial approximant, the second to zero (with lengthening of the preceding vowel to maintain the metrical length of the syllable). The closest thing to a Latin -bd- comes at the rather weak morpheme boundary between the prefix ab- and words beginning with d-: the result is simply -bd-, as in abdico, abduco, etc. As far as I know neither -gd- nor -zb- (or -sb-) ever occurred in any stage of native Latin; -gd- was tolerated in borrowings from other languages, like amygdalum “almond.”

OPD also makes, without providing examples, the following statements about Telerin:

1) *ŋʲ- > j- initially (OPD p. 32)

2) *ŋʷ- > m- (OPD p. 32)

3) *kʷ- > p- (OPD p. 35)

4) ŋ is not found “in isolation” in Telerin (OPD p. 36)

5) Telerin had a strong stress accent, which at an early date was moved toward the beginning of the word (“retracted”) and placed on the first syllable of the stem, usually the first of the word unless prefixed. (OPD pp. 56-7)

(1) Is contrary to the development shown in Comparative Tables, where *ŋʲ- > g-.

A second version of the Tengwesta Qenderinwa (TQ2), published in Parma Eldalamberon #18, pp. 71-107, has no new Telerin words, only repeating the example tancula (p.100). But it does have some notes (without examples) on the development of Telerin sounds, which represents a departure from Comparative Tables.

According to this outline (pp. 103-105), the Quendian labialized velars became labials, a consistent trait of all stages of Telerin; palatalized velars became coronals, consistent with The Etymologies and differing from Comparative Tables; and, conversely, initial prenasalized stops became nasals, in agreement with Comparative Tables, but contrary to the evidence of The Etymologies and later works.

Thus,

Quendian *gʷ *ŋʷ *ŋgʷ *w = Comp. Tab. b m m v = TQ2 b m m v

Quendian *gʲ *ŋʲ *ŋgʲ *j = Comp. Tab. g g g j = TQ2 d n n y

Quendian *g *ŋ *ŋg *ȝ = Comp. Tab. g g ŋg (sic) g = TQ2 g ŋ ŋ h

Quendian *b *mb = Comp. Tab. b m = TQ2 b m

The correspondences *ŋ = T. ŋ, and *ȝ = T. h are worth noting, since they would be contradicted or modified at a later date.

It is also stated that *-ŋŋ- > -ŋg-, *-ŋʲ- > -ŋgʲ-, *-ŋʷ- > -ŋgʷ- were Telerin medial developments — and, although it is not clearly stated, it can be deduced that these two shifts took place prior to the elimination of the Telerin palatalized and labialized velars. The normal developments are then *-ŋʲ- > *-ŋgʲ- > -nd(i)- and *-ŋʷ- > *-ŋgʷ- > -mb-. These medial changes could not have taken place afterwards, since the depalatalization of *-ŋʲ- and the labialization of *-ŋʷ- would have produced **-n-, **-m-, which would bleed the stated rule, depriving it of any possible input.

ŋ is also supposed to have remained between vowels. The resulting presence of ŋ- and -ŋ- in Telerin would contradict the statement on OPD p. 36 that isolated ŋ did not exist in Telerin — and, in fact, no Telerin words with initial or intervocalic ŋ are exemplified anywhere.

Pettontāli

Although I have used some examples from The Etymologies to illustrate some of the characteristics shown in Comparative Tables, the Telerin of Comparative Tables is not identical with that of the The Etymologies. In fact, very important and lasting changes took place between the two. These changes do not, however, lessen the resemblance of Telerin to Latin.

In Comparative Tables, the distinctive series of Quendian palatalized velars (kʲ, kʰʲ, gʲ, ŋʲ) is eliminated in Telerin by a wholesale merger with the plain velars. In The Etymologies, the palatalized velars are still eliminated, but by a merger with the coronals, as in TQ2. This is shown by the example already mentioned of *ŋ̩gyo > endo “grandchild”; according to Comparative Tables, it would have been ango (or perhaps rather engo). Another example is perhaps the most famous word in Telerin, the reflex of *kʲelepē — given in The Etymologies as telpe “silver.” It is famous because it was borrowed into Quenya (displacing the earlier, regular reflex tyelpe, and so became the base for various notable names, like Telperinquar “Celebrimbor” (cf. Unfinished Tales, p. 266). In the phonology of Comparative Tables, the word would rather have been **cel(e)pe. Other examples are Fallinel and Solonel, two epithets for the Teleri, where the element -nel (-nell-, -neld-) comes from a root !ŋʲel- (LR:379 NYEL-; The Etymologies uses the spelling ny for the palatal nasal). The words are compounds, containing !nel(lo), pl. nelli “singers,” and therefore internal -ni- is not seen.

More briefly:

1) Comp Tab.: *kʲ *kʰʲ *gʲ *ŋʲ *ŋgʲ > c h g g g

2) TQ2: *kʲ *kʰʲ *gʲ *ŋʲ *ŋgʲ > (t) (?) d n n

2) Etymologies: *kʲ *kʰʲ *gʲ *nʲ *ŋgʲ > t (f) (d) n (d)

Another major change was in the treatment of the initial sequences of nasal+voiced stop. In Comparative Tables, they become nasals for the most part, though ŋgʲ is shown becoming g, and ŋg is shown (inconsistently and improbably) remaining ŋg.

In The Etymologies and later works, however, they are voiced stops, as in the word golodo “Noldo” from *ŋgolodō.

Thus:

1) Comp. Tab.: *mb *nd *ŋgʲ *ŋg *ŋgʷ > m n g ŋg (sic) m

2) TQ 2: *mb *nd *ŋgʲ *ŋg *ŋgʷ > m n n ŋ m

2) Etymologies *mb *nd *ŋgʲ *ŋg *ŋgʷ > (b) (d) (g) g (b).

The values in parentheses are speculative (though most are exemplified elsewhere), due to the paucity of data that The Etymologies provides. Despite this, more information on Telerin phonology can be milked from the text. The innovations, or, conversely, the conservative tendencies that differentiate Telerin from Quenya are shown below:

Innovations

1) *eu > ū (Q. eu) : T. būa, būro < *beujā-, *beurō (LR:352 BEW-)

2) *ŋ (syllabic) > en (Q. in): T. endo, Q. indyo (LR:377 ÑGYO)

3) *kʷ > p (Q. qu): T. alpa, Q. alqua (LR:348 ÁLAK-)

4) *θ > f / #_ (Q. s): T. findo, Q. sindo (LR:392 THIN-)

5) *θ > d / V_V (Q. s): T. bredele, N. brethel, brethil (LR:352 BERÉTH)

6) *gʲ > d (Q. y, dy): T. endo, Q. indyo (LR:377 ÑGYO-)

7) *ŋg > g / #_ (Q. ŋ > n) : T. golodo, Q. noldo (LR:377 ÑGOLOD-)

8) *nʲ > n (Q. ny): T. Fallinel, Solonel, Q. Falanyel, Solonyeldi (LR:379 NYEL-, 381 PHAL-, 387 SOL-)

9) *j > i / C_V (Q. y): T. spania, Q. fanya (LR:387 SPAN-).

Most of these changes have already been noted. The change (1) of *eu > ū is, however, different from that in Comparative Tables, where *eu > iu (Parma Eldalamberon #19, p. 25). This change is also found in Latin, where *eu > ou (attested in inscriptions) > ū, e.g. *deukō > doukō (douco) > dūkō (duco) “I lead.”

Conservative tendencies

1) *b = b (Q. v): T. bala, Q. vala (LR:350 BAL-); so also Bana, Baradis, belda, belka, belle, bredele, būa, būro.

2) *g = g (Q. zero, y): T. felga, ulga, Q. felya and cf. Q. ulundo (LR:381 PHÉLEG-, LR:396 ÚLUG-)

3) *sp = sp (Q. f): T. spania, spanga, spalasta-, Q. fanya, fanga, falasta- (LR:387 SPAN-, SPÁNAG, 381 PHAL-)

4) *st = st (Q. s, N. th): T. stalga, N. thala (LR:388 STÁLAG-)

5) *w = v (Q. w): T. vilverin, Q. wilwarin (LR:398 WIL-)

6) Telerin shows syncope of the 2nd vowel of a CVCVC root sequence less often than Quenya. T. Baradis, golodo, Q. Varda, noldo. Baradis is probably *Barada ( = Varda) + dis(se) = Q. nisse ‘woman’. (LR:351 BARÁD-)

A scrap of paper associated with The Etymologies (published in Vinyar Tengwar #46, pp. 28-29) identifies a plural element -m (from a root √eme- ‘many’) which appears in Telerin as -am, -um, or -em. The example cited is edulam, presumably the plural of !edula; given the example of tancula already mentioned, we can assume that edula reflects earlier *edlā, and is therefore cognate to Q elda and N edhel; i.e. Telerin edulam = Q. eldar.

This -m suffix vanishes from Telerin, never to be seen again (at least as a plain plural marker), but it does at least show that Telerin, like Latin and unlike Quenya, at this time tolerated a final consonant -m.

Lambecentai

Additional information on Telerin is, happily, available in other works that allows us to build up a more complete view of Telerin phonology, and (to a much smaller extent) its grammar. These works are:

*The Outline of Phonology (OP), published in Parma Eldalamberon #19, pp. 69-107, probably dating from the early 1950s (as evidenced by the use of the term ‘Sindarin’).

*The essay Quendi and Eldar, published in The War of the Jewels, pp. 360-417, except for a section published in Vinyar Tengwar #39, pp. 4-17, dating around 1959-1960.

*An essay on the names of the rivers and beacon-hills of Gondor, parts of which were published in Unfinished Tales (in which, however, the only remarks about Telerin are on pp. 265 and 266), and the remainder in Vinyar Tengwar #42, pp. 8-27. This essay dates to c. 1969.

*An essay called by Christopher Tolkien The Shibboleth of Feanor, part of which was published in The Peoples of Middle-earth pp. 331-366, and the remainder in Vinyar Tengwar #41, pp. 7-10, to which is related an incomplete essay on the Quenya word ōre, also in Vinyar Tengwar #41, pp. 11-15. These date from some time after 1968.

*A group of essays on Eldarin words for the hand (in different positions), the fingers, and the Eldarin numbers, published in Vinyar Tengwar #45, pp. 6-47 and Vinyar Tengwar #48, pp. 4-11, 24-26, 30. These also date after 1968.

A few other words or notes about Telerin appear in scattered locations, which will be cited as needed.

These works are not entirely consistent with each other — even those which were published in the same time frame — and in some places show what may be either a reworking of Telerin phonology, or casual errors, especially those introduced by a confusion between Telerin and closely related forms (e.g. in Quenya).

Lāmapenta

The Outline of Phonology is a rewriting of the Outline of Phonological Development, and is structured along very similar lines. References to Telerin occur at the same place as in OPD, but the facts presented are different. The Telerin words cited in OPD are resta “sown field, acre,” hacta “fence, hedge,” nacte “wound,” necte “honey,” hamna “pile, mound,” hauda “hoard, store” (OP p. 91), buspo “large fly,” and mascu “mire” (OP p. 101). These represent Telerin reflexes of the Quendian consonant clusters -dd- (-zd-), -gd-, -bn-, -wd-, -zb-, -zg-.

The phonological data revealed are:

Telerin innovations

1) *-gd- > -ct- (Q. -ht-) T. hacta, Q. hahta, T. necte, Q. nehte, T. nacte, Q. nahte < *kʰagdā, *negdē, *snagdē.

2) *-zb- > -sp- (Q. -ps-): T. buspo, Q. pupso < *buzbō. Total devoicing of the whole cluster.

3) *-zg- > -sc- (Q. -ks-): T. mascu, Q. makso < *mazgu.

(1) -gd- does not seem to occur in Latin, but the sequence -gt- produces -kt-: *agtos > *aktos > aktus (actus) “done.” In this version of Telerin, -ct- has an alternative form -tt-: beside hacta, necte, nacte we also have hatta, nette, natte. This is like the development from Latin to Italian: octo > otto “eight,” noctem > notte “night.” The developments above are again in contrast to that in OPD, where *kʰagdā > hāda, not hacta.

(3) A different development, -zg- > -rg- occurs in one Latin word, *mezgō > mergō “dip.” Otherwise -sg- usually > ːg, e.g. *disgesō > *dizgezō > dīgerō (digero) “I arrange.” Note also that a final *short* u remains in Telerin, whereas in Quenya it would > o. Latin final -u is normally [ū], but neuters like cornu [kornū] would seem to reflect older forms like *kṛnu (with short final vowel).

Innovations shared with Quenya

1) *-dd- > *-zd- > -st- (Q. same): T., Q. resta < *rezdā < *reddā

2) *-bn- > -mn- (Q. same): T., Q. hamna < *kʰabnā

3) *kʰ- > h- (Q. same): T. hacta, hamna, hauda, Q. hahta, hamna, haura

4) *sn- > n- (Q. same): T. nacte, Q. nahte < *snagdē

(1) Is not a Latin development; -dd- did not occur word-internally, and between morphemes remained -dd- (e.g. adduco). -zd-, as mentioned, did not produce -st- but rather -ːd-. This is a change in process from that in OPD, where -zd- > -rd-.

(2) Parallels both Quenya and Latin: for the latter, cf. *sabniom “land of the Sabines” > Samnium, *swepnos > *swebnos > *swemnos > *swomnos > *somnos > somnus “sleep.”

(3) *kʰ > h also occurred in Latin, where I-E *gʱ > *kʰ > *x > h initially (*gʱostis > *kʰostis > *xostis > hostis “enemy”).

(4) *snagdē > nacte exemplifies the change of initial sn- to n-. This parallels a Quenya change, but also is found in Latin: *snigʷʱs > *snikʰs > niks (nix) ‘snow’. It is mentioned on p. 79 of OP that in all Eldarin dialects (including Telerin) initial *sm-, *sn- change to m-, n-.

Conservative tendency

1) *-d- = -d- (Q. -r-): T. hauda, Q. haura < *kʰaudā

(1) -d- is preserved in Latin, as in laudō “I praise.”

 

OP also includes some unexemplified statements about innovative phonological developments in Telerin:

1) *h- > zero (OP p. 74)

2) *ŋʲ > ŋgj (Q. ny-, -ndy-) (OP p. 74)

3) *ŋʷ > ŋgʷ (Q. ŋʷ-, -ŋgʷ-) (OP p. 74)

4) *kʲ > *tʲ > t (Q. ty) (OP p. 75)

5) *gʲ > *dʲ > d (Q. ly) (OP p. 75)

6) *mb-, *nd-, *ŋg- > b-, d-, g- (Q. m-, n-, ŋ-) (OP p. 76)

7) *-mt- (and *-md-?) > -mpt- (OP p. 98).

8) *-ln- > -ll- (OP p. 100)

9) *-ȝs- > -ss- (OP p. 74)

(1) In this description, *h is apparently an alternative to *ȝ, not a different phoneme. This account of its disappearance contradicts the account in TQ2, where *ȝ- > Telerin h, and also contradicts subsequent developments, where two distinct phonemes show different developments in Telerin: *ȝ- > zero, but *h- > h-.

(2,3) As in TQ2 p. 104, except that here initial sequences *ŋʲ-, *ŋʷ- also > *ŋgʲ-, *ŋgʷ- in all Eldarin dialects. They would yield Telerin *ndʲ-, *mb- > d-, b-. But in light of the continuing correspondence *ŋʲarnā > Q. nyarna, S. narn, the first development at least does not seem to be sustainable; rather, *ŋʲ- should have > *nʲ- > T., S. n-.

(6) is the normal development henceforth.

(7) Cf. Latin *emtos > emptus “bought,” *sūmtos > sumptus “taken up”; though this is a late and unusual development, as normally *-mt- > -nt-.

(9) Seems to be stated to be a Common Eldarin change, and an example is found in the Noldorin of The Etymologies, *waȝsē > ON. wasse, N. gwass. But there Q. has the development *-ȝs- > -ks-, whereas in OP Q. also has *-ȝs- > -ss-.

It is also noted that “the initial groups dr, gr, gl were favoured in Telerian” — referring, however, to the entire branch of Telerin languages, including Nandorin and Sindarin, and not to the Telerin of Aman specifically. (OP:80)

Pendi ar Elloi

The next large batch of Telerin words comes from the wide-ranging essay Quendi and Eldar. The following phonological elements are exemplified in them, page citations being from The War of the Jewels unless otherwise specified.

Innovations

1) *ǣ > ā (Q. ē): T. māla, Q. mēla “loving, affectionate” (Vinyar Tengwar #39, p. 10).

2) *ō (originally final) > u (when medial): T. Hecellubar (<Hecello+bar), Q. Hekeldamar (pp. 365, 376)

3) *ŋ (syllabic) > (Q. in): T. engole, Q. iŋgole “lore” (p. 383).

4) *l > ul/C_(V)# (Q. il): T. hecul or heculo, Q. hekil, hekilo (p 365).

5) *kʷ > p (Q. qu): T. epe, Q. eque “says, said” (p. 392), T. abapētima, Q. avaquētima “not to be said” (p. 371), T. pen, Q. quen “one, somebody, a person” (p. 362), T. aipen, Q. aiquen “whoever,” T. arpen, Q. arquen “a noble,” T. arpenia (*arkʷenjā) “noble,” T. ilpen, Q. ilquen “everyone,” T. pendi, Q. quendi, “elves (esp. of the earliest period of their history),” T. calapendi, Q. kalaquendi “light-elves,” T. moripendi, Q. moriquendi “dark elves,” archaic T. pendia, Q. quenya “language of the Quendi” (p. 375).

6) *mb > b/#_, +_ (Q. m): T. -bar, Q. -mar “land, dwelling” (pp. 365, 376).

7) *ŋg > g/#_ (Q. ŋ > n): T. goldo, Q. ŋoldo > noldo “one of the Noldor,” T. gōle, Q. ŋōle “study” (p. 383).

8) *ld > ll (Q. ld): T. ella, ello, Q. elda, eldo (pp. 362, 375-6), T. ellālie, Q. eldalie (p. 375), T. galla, Q. alda “tree” (Vinyar Tengwar #39, p. 7).

9) *dl > ll (Q. ld): T. ello, Q. eldo (p. 364).

10) *ln > ll: T. delle (< *delnē) (p. 364).

11) *j > i / C_V (Q. y): T. delia-, Q. lelya- “go” (p. 364), T. vania, Q. vanya “one of the Vanyar” (p. 383), T. cavaria, Q. coarya “his house” (p. 369), T. pendia, Q. quenya “language of the Quendi,” T. arpenia “noble” (p. 375), Q. not found but would be !arquenya.

12) *w > u / C_V (Q. w): T. Olue, Q. Olwe, name of the King of the Teleri (p. 369); T. vanua, Q. vanwa “gone away” (p. 367).

13) *s > r / V_V (Q. same): T. cavaria “his house” (<*kawasjā), Q. coarya (p. 369).

14) *sd > *zd > ːd (Q. st): T. ēde, Q. este “the Valie Estë” (pp. 403-4).

Many of these changes have already been seen, and show the continuity that (uniquely among Tolkien’s ‘minor languages’) characterized Telerin.

Among the new changes are:

(1) *ǣ > ā. This *ǣ (a long, low front vowel, derived from a-infixion in a root with sundóma E) is a vowel peculiar to Primitive Quendian (lost in all descendent languages) and only deducible from pairs such as T. ā/Q. ē. A similar *ǭ would also have given T. ā/Q. ō, though a word of this form is not directly attested; we can reconstruct T. !cālo “burden” by comparing Q. kōlo, S. caul. Cf. Vinyar Tengwar #39, p. 10. These developments notably already appeared in Comparative Tables and were clearly never lost sight of.

The development (2) o > u is somewhat mysterious. It is evidently not the reflex of an original *-u, as hecello includes as its second element ello < *edlō. In other compounds, medialized -ō appears as -o-, e.g. goldolambe “language of the Noldor.” In Latin, a conjoining -o- generally became -e-, then -i-: e.g. *signo-foros > *signeferos > signifer “standard-bearer,” *auzo-faks > *aurefeks > aurifeks (aurifex) “goldsmith.” But -i- sometimes became -u- preceding a labial consonant (p, b, f, m); e.g. aurufex beside aurifex, and regular volumus “we want” for **volimus. Perhaps Hecellubar comes from older *hekellobar > *hekellebar > *hekellibar, with the subsequent change of i > u conditioned by the following b of bar. The main problem with the idea is that other examples of vowel reduction, similar to those of Latin, are not seen in Telerin.

(4) heculo shows the same change as tancula, but the Quenya derivatives are different; Quenya had been altered so that the svarabhakti vowel between a consonant and l before a final vowel (or itself final) was i rather than a; other examples include *tekla > tekil “pen,” *makla > makil “sword,” presumably = T. !tecul, !macul.

(5) The Telerin change of *kʷ > p, which was part of an extensive change that converted the entire Quendian system of labialized velars into labials, is supposed to have taken place very early in the history of the Lindar, the ‘third branch’ of the Quendi from which the Teleri emerged. It had already taken place in Common Telerin, i.e., the language ancestral to the Telerin of Aman, to Sindarin, and to the languages of the Nandor (The War of the Jewels, pp. 375, 407). It may be even older, and have taken place even among the Lindar of Cuiviénen, prior to the separation of the Eldar and the Avari (The War of the Jewels, p. 410).

(8, 9, 10) The changes *ld, *dl, *ln > ll are found in Latin: *sedlā > sella “stool” (root *sed- “sit”); *saldō > sallo “salt” (rare; more usually sal); *kolnis > kollis (collis) “hill”; the same also in Telerin, by way of *-ln- > Common Eldarin *-ld- > -ll- (OP p. 100); but Telerin -ln- arising by syncope appeared later and remains: *elenī > elni “stars” (The War of the Jewels pp. 360, 362).

Some inconsistency appears in the development of -ld-. It is generally > -ll- (*eldā > ella, *galadā > *galda > galla), but we also consistently have goldo for *ŋgolodō, not **gollo. Possibly the intention was original *-ld- > -ll- , while -ld- arising from syncope > -ld-; but then we should have *galadā > galda. In both the earlier Etymologies, and in the much later Shibboleth of Feanor, we find golodo, without syncope; though in that case, “tree” probably ought to be galada, not galla or galda.

The change *dl > ll carves out an exception to the general rule that an epenthetic -u- is inserted between obstruent and l, which was seen in tancula and edulam, and is still seen in heculo. The exact parameters of the exception are uncertain; it could apply only to *dl, or to all stops+l other than *kl, or to some intermediate set. In Latin *dl > ll, but *tl > *kl (merging with original *kl) > kol > kul, *pl > pol > pul , and *θl > *fl > *fol > *vol > *bol > bul (*pōtlom > poculum, *fakl > facul, *poplos > populus, *staθlom > stabulum). *sl however > *zl > ːl (*preslom > prēlum prelum). When the l in a cluster of obstruent+l preceded an i (short or long), it developed an epenthetic i preceding it: *faklis > facilis, *staθlis > stabilis. This was due to the l in such positions being slightly palatalized; the resulting l exilis was syllabified as il, while the remaining non-palatalized l was syllabified as ul.

(12) generalizes the change, already seen, of post-consonantal *j > i to one in both the glides *j, *w are syllabified. This sometimes occurred in Latin: e.g. *kʷetwōr > *kʷattwōr > kʷattuor (quattuor) “four,” *mṛtwos > *mortwos > *mortuos > mortuus “dead,” *tenwis > tenuis “thin.” But it did not normally occur after l; a Latin **olwe- would have > **olle- or **olve-, depending on the time of the appearance of the -lw- cluster; cf. earlier *moldwis > *molwis > mollis “soft,” later *salowos > *salwos > salwus (salvus) “safe, well.”

(13) was one of the most common and characteristic changes in Latin: *esō > erō (ero) “I will be,” cf. esse “to be”; *eusō > *ousō > ūrō (uro) “I burn,” cf. ustus “burnt”; *temposes > temporis “of time,” nom. tempus, *mūses > mūris (muris) “of a mouse,” nom. mus; *koisā > kūra (cura) “care”; *ausom > aurum “gold.” The change is thought to have been via through an intermediate *z (i.e. s > *z > r), but although the change took place in historical times, *z can only be postulated by indirect evidence; until the merger with /r/ took place, [z] was only an allophone of /s/, and was always written with the letter s.

As in Quenya, Telerin *s remained s (or rather changed > *z > s), even intervocalically, when the consonant before or after the vowel adjoining *z was r. Examples are known only from Quenya: *kazar > kasar “dwarf,” a borrowing from Khuzdul kʰazād (The Peoples of Middle-earth, p. 388), *raze > rase “sticks out,” from the root √ras- also seen in rasse “horn, peak”; but they are asserted to have taken place in Telerin as well (OP p. 73). This change, or lack of change, is also found in Latin: *miseros > *mizeros > miser “wretched,” not **mirer.

(14) Shows the final development of -zd-, after two previous experiments: *ezdē > erde, *reddā > rezdā > resta, and finally *ezdē > ēde. The last brings it into line with Latin nīdus (nidus) < *nizdos by compensatory lengthening; but it is possible that the intended development was *-z- > *-i-, i.e. *ezdē > *eide > ēde. The correspondence *ei > ē is noted elsewhere (Vinyar Tengwar #48, p. 7).

Conservative tendencies

1) *b = b (Q. v): T. aba-, Q. ava- “refuse,” T. abā, Q. avā “don’t,” T. abapētima, Q. avaquētima “not to be said,” T. abari, Q. avari, “refusers, Avarin Elves.”

2) *d = d (Q. r, medially and finally): T. audel, Q. aurel, “elf who went to Aman,” T. avad, Q. oär “away.”

3) *g = g (Q. zero): T. galla, Q. alda “tree.”

4) *h = h (Q. h, S. zero): T. heca, Q. heka, S. ego “begone!”

5) *kt = ct (Q. ht): T. hecta-, Q. hehta- “abandon.”

6) *w = v (Q. v, zero before o): T. Vania, Q. Vanya “one of the Vanyar,” T. vo-, -, Q. o-, ō- “together,” T. vomentie, Q. omentie “meeting.”

7) Stressed medial long vowels not shortened: T. ellālie, Q. eldalie “elven-folk,” T. goldōrin, Q. ŋoldorin “of the Noldor,” T. lindārin, Q. lindarin “of the Teleri.”

8) Archaic syncope maintained: T. elni pl. of elen “star,” Q. eleni was reformed after elen.

9) Maintenance of *w as a consonant between vowels: T. cava, Q. coa “house,” T. avad, Q. oär “away.”

With respect to (7), shortening of an original long vowel does occur in several other compounds: goldo-lambe, linda-lambe, cala-pendi, and the later-attested alatāri(g)elle, findarāto, angarāto. These probably derived from *goldōlambe, *lindālambe, *calāpendi, *alatārīgelle, *findārāto, *angārāto. (The forms *findārāto, *angārāto would have arisen by coalescence of a final vowel and following initial vowel; cf. *kirja-aran > *kirjāran > ciriāran “ship-king.”)

The implicit rule is that a long vowel is shortened when immediately followed by a heavy syllable — one that is either long by nature (with a long vowel or diphthong) or long by position (i.e., in a syllable closed by two or more consonants). If two heavy syllables are adjacent, the first of the two, if containing a long vowel, is shortened. The process works from right-to-left (i.e., from the end of the word to the beginning), so that in *alatārīgelle, the heavy (by position) penultimate syllable gel prompts the shortening of the vowel of the previous syllable ; this now being short (*alatārigelle), the preceding syllable is not further shortened. This type of shortening is very suggestive of a Latin-type main stress on a penultimate or antepenultimate syllable, the shortened vowels all occurring in presumably unstressed syllables prior to a stressed penultimate; but it is fair to admit that nothing explicit has been published regarding stress in Telerin.

In the compound ellālie (ella “elf” + lie “people”), the penultimate vowel is short, and therefore the vowel of the previous syllable is preserved; thus we do not see **ellalie. In hecellubar, the final syllable evidently counts as heavy (perhaps *hekellōbār > *hekellobār > hecellubar) so we do not get **hecellōbar, but in goldōrin, lindārin the suffix -rin is evidently considered light; likewise the final syllable of ciriāran.

A few remarks are in order about the morphology of Telerin revealed in these words.

We do not find Telerin plurals in -m, like edulam, but we find many of them in -i, where Quenya has both -i and -r: ēli, elni “stars” (Q. eleni), Abari (Q. Avari), Audelli (Q. Aureldi), Calapendi, Moripendi (Q. Calaquendi, Moriquendi), Vaniai (Q. Vanyar), Lindai (Q. Lindar), Elloi (Q. Eldor, Eldar), Hecelloi (Q. Hekeldi).

Telerin was inflected like Quenya, but with differences in detail. Thus the genitive marker in both was -o, but while the Quenya genitive plural was -on, in Telerin it remained -o. Thus Q. Avarion “of the Avari,” but T. !Abario of the same meaning. The doubling of the consonant in the suffixes -llo, -sse-, -nna seems not to have occurred in Telerin; thus Q. lūmenna “on the hour” = T. lūmena, i.e. more transparently lūme “hour”+na(ː)”to.” The shortening of the e in *lūmēnā points to an originally final stress; otherwise we would expect **lumēna.

Possessive suffixes were in use, including -ria “his” (Q. -rya) and -ngua “our (two)” (Q. -lva – an inclusive form, either dual or indefinite between dual and plural at the time Quendi and Eldar was written). The ending -ngua cannot be plausibly derived from Quendian *-ŋʷa or *-ŋgʷa, as those suffixes, if present in Quendian, would have produced **-mba in Telerin. Nonetheless, -ngua must reflect an immediate *-ŋgʷa as its precursor, with w > u as in Olue, vanua; but that same *-ŋgʷa must have arisen after the transformation of Quendian *-ŋgʷ- to -mb- in Telerin. Unfortunately we don’t know exactly what sound-combinations in Telerin would produce -ŋgʷ-, and the Quenya pronominal paradigms, in terms of both form and meaning, are an exceptionally tangled maze.

However, for the stage of development of the Eldarin languages reached at the time of Quendi and Eldar, I think a plausible antecedent is *-ŋkma, a termination whose constituent elements are n(i)- “1st person singular,” k(i)- “2nd person familiar,” and m(e)- “1st person plural” (at the time, without reference to the numbers involved or the inclusion/exclusion of the person addressed). The combination of n- and k- to produce a first person inclusive -ŋk- is well attested from both Quenya and Sindarin verbal and possessive pronominal endings (Parma Eldalamberon #17, p. 46 (S. lammenc 1st incl. pl.), p. 57 (Q. cariŋgwe and carinque 1st. incl. dual), p. 75 (Q. -ngwe 1st incl.pl., -ngo 1st incl. du.) p. 132 (S. cerinc, galanc, 1st incl. du. & pl.), Vinyar Tengwar #49, p. 51 (pre-Quenya -inke 1st incl. du.)). The Quenya 1st person dual/plural verb endings are formed by adding suffixes -me, -mo, -we, -wo (distributed according to different principles at different points in the history of the inflection). -inque certainly reflects -i+ŋk+we; -iŋgwe very plausibly is -i+ŋk+me.

With regard to the phonology, [k] would be voiced to [g] by partial assimilation to the following nasal, producing *-ŋgma (the voicing of -km- > -gm- is said to be Common Eldarin, PE #17 p. 43), which would then likely simplify to *-ŋm-; [m] could then be dissimilated to [w], an unusual change but one that occurs frequently after nasals in Quenya. The subsequent change *-ŋʷ- > *-ŋgʷ- is a natural epenthesis, already said to have occurred at an earlier stage in Telerin (OP p. 74), and frequent in Quenya: cf. *teŋʷē > teŋgwe in The War of the Jewels p. 394, and the complex Quenya sound-change *-km- > *-gm- > *-ŋm- > *-ŋʷ- > -ŋgʷ- mentioned in PE #17 p. 44. Since in Quenya at this stage *-ŋʷ-, *-ŋgʷ-, *-km-, and *-ŋkm- all fell together as -ŋgw-, while in Telerin -ŋʷ- and -ŋgʷ- instead merged with -mb-, Telerin -ngua is — despite initial appearances — firm evidence that the second element in Q. -iŋgwe was -me and not -we.

Latin had similar, though not identical changes. Latin *k > *g before nasals: *sekmn̩tom > *sekmentom > *segmentom, *deknos > *degnos. The resulting g, and original *g, then > ŋ: *degnos > *deŋnos > *diŋnos > diŋnus (dignus) “worthy,” *legnom > *leŋnom > *liŋnom > liŋnum (lignum) “wood”; thus merging with *ŋgn > ŋn: *iŋgnoskō > iŋnoskō (ignosco) “I excuse.” There was no change of m > w, however; the resulting clusters were -ŋm-, -ŋn-, spelled (and later pronounced) gm, gn.

The verbal inflections were similar to Quenya. Past tenses include delle (*delnē); vāne (root √wa-: + suffix -) “vanished, departed, was lost,” vante “went to another place,” probably from avante (stem *awa-t(a)- with n-infixion + -ē) remodelled after vāne; cf. Q. oante. Perfect forms show reduplication are avānie “has departed, has been lost” and avantie “has gone elsewhere.” The first person ending is -n (aban “I refuse”), the imperative ending is -a (ela “behold!,” heca “scram!”)

The notes (from very disparate sources, for the most part written shortly before Quendi and Eldar) regarding the invented words and names in The Lord of the Rings, published together in Parma Eldalamberon #17 as Words, Phrases and Passages, have almost nothing to say about Telerin. One form, nimbi “white,” is cited three times (pp. 19, 49, 168) and twice called “Telerin”; but in both cases, “Telerin” may not mean “the language of the Amanya Teleri” but rather “belonging to the Telerin branch of Eldarin,” i.e., Proto-Telerin or even Proto-Sindarin. This is suggested by the archaic retention of a final short -i. Regardless of the meaning, the form is odd; it clearly has a relationship to Q. ninque “white,” which in turn relates to the root √nikʷ- “cold” (presumably via the whiteness of snow, ice, or frost), but it is differentiated from *nimpi, the expected Telerin reflex of Quendian *niŋkʷi, which lies behind the niph- of niphredil, and which seems to have the basic meaning “pale, pallid.” The reason for postulating a distinct nim and nimp (in Sindarin), however seems to be to explain why the Ered Nimrais were not the **Ered Niphrais and Nimrodel was not **Niphrodel.

In addition to the citations of nimbi, the suggestion is made, in a note included in Words, Phrases, and Passages, that Altarielle (i.e., Galadriel) is actually a Telerin name, distinguished from a Quenya counterpart with initial ŋ-. The idea is propounded, without elaboration, that *ŋ- > zero in Telerin. (PE #17, p. 60) — contrary to both Comparative Tables and the second Tengwesta Qenderinwa. This idea, however, would prove to be lasting, and will be dealt with below.

Endoriai Olue

Most of the latest data on Telerin comes from the essays of the late 1960s. Unfortunately, rather than showing a greater coherence of conception, a certain amount of contradiction, even of long-established concepts about Telerin, emerges. There are conceptual contradictions not only between texts (whose exact chronological relationship is uncertain) but even within the same text. Thankfully, the points involved are fairly minor, but the inconsistencies do complicate our understanding of the language.

The essay The Shibboleth of Feanor and its associated writings is primarily concerned with minor phonological details of the Quenya of the Noldor, and with the names of the principal lords and ladies of the Noldor. The presence of Telerin in these writings is thus oblique, involved mostly in explaining the names of the children of Finarfin by Eärwen of Alqualonde: Finrod, Angrod, Aegnor, and Galadriel. Together with Shibboleth is an unfinished essay on the word ōre, which contains a short Telerin phrase.

The phonological details revealed in this essay are for the most part familiar. Citations below are from The Peoples of Middle-earth (PME) and Vinyar Tengwar #41 (41).

Innovations

1) *kʷ > p (Q. qu): T. pete, Q. quete “speaks” (41 p. 11)

2) *kʲ > t (Q. ty): T. telpe, Q. tyelpe “silver” (PME p. 356)

3) *g > zero/ V_V (Q. same): T. rīa, Q. rīa, T. rielle, Q. rielle (PME p.347).

4) *ŋg > g/#_ (Q. ŋ > n): T. golodo, Q. ŋoldo (PME p. 360).

5) *ŋ > zero/#_(Q. ŋ > n): T. alata, Q. ŋalta (PME p. 347).

6) *ȝ > zero/#_ (Q. same, S. g): T. ōre, Q. ōre, S. gūr ‘heart, spirit’ (41 p. 11).

7) *j > i/ C_V (Q. y): T. ciria, Q. cirya “ship” (ciria in the name Ciriāran, sc. ciria+aran “ship-king,” PME p. 341); T. -nia, Q. -nya “my” (41 p. 11).

8) *j > i/ V_V (Q. y): T. gāia, Q. āya “terror,” T. gaiar, Q. ear “sea” (PME p. 363).

9) C > zero / C_# (Q. same): T., Q. urus “copper” < *urust- (41, p. 10).

(3) is a very problematic development for Telerin. General *g > zero is a Quenya, not a Telerin development; in Telerin, medial -b- and -d- are preserved unchanged (an example of medial -d- in the verb glada- occurs in this very text) and -g- appears in a word loga in the “Rivers” essay (Vinyar Tengwar #42, p. 10). Eldarin *rīgā would be expected to give T. rīga. The only way to reconcile the forms loga and rīa as simultaneously Telerin is to assume that g is treated differently after i, which as a rule is suspiciously ad hoc. It is also unlike Latin, where internal g is preserved after any vowel; *jugom > jugum (iugum) “yoke,” *srīgos > frīgus (frigus) “chill.” Another possibility is that T. rīa is actually a loan from Q., but such a loan seems unmotivated. Appropriate forms could be generated from *rīȝā, but among the several competing etymologies of Galadriel’s name, the one constant is that it contains a root √rig-. Alatārielle could be an accomodation to Quenya phonology, but Tolkien insists that it is Telerin. I think Tolkien postulated rīa less with a view to its suitability to Telerin’s phonetic history than to justify the statement that Galadriel’s Telerin name was Alatārielle, and to avoid the less euphonious (to his ear) **Alatārigelle. Alatārielle is quoted as having the by-form Alatāriel. The fact that the -i- is still short (not **Alatarīel) suggests either that the shortening took place before the loss of a final vowel, or that we are to understand an underlying /alatārī(g)ell/ with a heavy final syllable, before which the vowel was shortened.

(4) shows a different reflex of *ŋgolodō than that of Quendi and Eldar, golodo for goldo. This form is a reversion to that of The Etymologies, and also avoids the problem of expected goldo > **gollo. It does raise the question, however, of where syncope may occur in Telerin, and whether galla “tree” ought not actually be !galada.

(5) and (6) are new, and directly contradictory to statements made in the second Tengwesta Qenderinwa. They show that the development of Quendian initial *ŋ and *ȝ (probably a voiced velar fricative) in Telerin was deletion (in TQ2 they became ŋ- and h-, respectively). They also point to a distinction between *ȝ and *h, although in earlier phonologies the two are allophones of a single phoneme. It might be supposed that Tolkien had simply changed his mind about Telerin reflexes of initial *h/*ȝ, i.e., that it was deleted instead of becoming h-; but another essay (published in Vinyar Tengwar #48, p. 26) is explicit (though without giving examples) that there were three sounds *ȝ, *h, and *ŋ, of which only *ȝ and *ŋ were lost in Telerin. Tolkien had actually changed his mind, but about Quenya, not Telerin; whereas in Quendi and Eldar Quenya preserved initial *h-, he now decided that it would be deleted in Quenya.

(7) The element nia “my” is actually quoted as a separate word in the phrase in which it appears: ōre nia pete nin “my heart speaks to me” (41, p.11). This is odd for several reasons. First, it’s an attached suffix in both Quenya (ōrenya) and Sindarin (guren) in the same essay. Second, we’ve already seen a similar attached suffix in Quendi and Eldar, cavaria “his house.” Third, if *nʲ- were actually phonologically initial, it should have been depalatalized to n- and we should see ōre na. I conclude that we must read ōrenia for ōre nia. Nin incidentally shows that the first person stem ni- and dative suffix -n were exactly as in Quenya.

(8) Is somewhat difficult to interpret, owing partly to Tolkien’s fondness for providing his languages with orthographies that obscure the phonetic structure. Gāia could conceivably be understood as [gā.i.a], [gāi.a], [gā.ja] . The first of these, with three syllables, is least likely, though not impossible. The second complicates our picture of Telerin diphthongs; while the third implies that, at least between vowels, i may be written for consonantal [j]. That is in fact the case in Latin, where spellings like maior, cuius represent [maijor], [kuijus] (or [majjor], [kujjus]). Likewise Tolkien vacillated in his spelling of Quenya between, inter alia, māya, maia, āyan, aian, hāya, haiya, aia, aiya, vaia, vaiya; it seems that āya, aia, aiya all spell the same sound, but whether it is in fact supposed to be [ā.ja], [ai.a], or [ai.ja] is unclear. So too with Telerin; but in Telerin unlike Quenya, aia and āia represent distinct and contrasting sequences, from *aja, *āja respectively.

Tolkien on more than one occasion (WJ p. 400, PME p. 363, PE17 pp. 27, 149) attempted to derive Q. ear “sea” from a root √gaj- producing *gajar-, Q. ear, S. gaear, T. gaiar. This theory, which had the advantage of tying several originally distinct stems together, unfortunately falls foul of the fact that in all published works Tolkien had used or cited Sindarin aear, aearon, in positions where it would not be lenited: e.g., in The Lord of the Rings, Oraearon next to Orgaladhad, Orgilion – a fact which Tolkien presumably overlooked. But in The Road Goes Ever On, the Sindarin word is cited as aear. The Adventures of Tom Bombadil likewise gives the name of a tower, Tirith Aear “guard of the sea,” with aear in the normally unlenited position of a modifying (genitive) noun. Q. ear must therefore be < *ajar-, and the Telerin word ought to be aiar. And although aesthetic considerations prove nothing, it seems to me preferable to think of King Olue of Alpalonde’s daughter, the spouse of Arafinue (Finarfin), as Aiaruen rather than Gaiaruen.

(9) The Telerin word for “copper” is quoted as urus, urust-; it shows that, like Quenya, Telerin reduced final consonant clusters (normally to a single consonant), e.g. *urust > urus; but the clusters remained when a suffix followed, e.g. genitive singular !urusto “of copper.” Such reduction happened also in Latin, but in a more restricted way due to the usual presence of overt inflectional endings, especially nominative -s; direct parallels are only found in a small number of words, usually neuters lacking inflectional endings in the nominative, such as *kord > kor (cor) “heart,” *dlakt > lak (lac) “milk,” genitives cordis, lactis.

Conservative tendencies

1) *pʰ > f /_u (Q. h): T. fuine, Q. huine, S. fuin “gloom” (41 p. 8)

2) *tʰ > θ (Q. s): examples not provided, but this is explicitly stated as a fact about Telerin (PME p. 332).

3) *d = d (Q. r or l medially): T. glada-, Q. lala- “laugh” (PME p. 359)

4) *g = g/#_ (Q. zero): T. gāia, Q. āya; T. gaiar, Q. ear (PME p. 363)

5) *gl = gl/#_ (Q. l): T. glada-, Q. lala- “laugh” (PME p. 359)

6) *w = v (Q. zero /_o): T. vola “long wave” (PME p. 357)

7) Glides *j, *w preserved between vowels as *i, *v (Q. *aya, *awa > ea, oa) : T. gaiar, Q. ear (PME p. 363); T. eve (PME p. 340)

8) Syncope avoided: T. alata, Q. ŋalta (PME p. 347); T. arāta, -arāto, Q. arta (PME pp. 346-7, 363); T. golodo, Q. ŋoldo (PME p. 360)

9) Medial long vowels retained long when stressed: T. Ciriāran (PME p. 341) , arāta, -arāto (PME pp. 346-7, 363)

(1) *pʰ > f occurs generally in all the Eldarin languages, but in Quenya f > h before a following u. This did not occur in Telerin, because Telerin f had early become a true labio-dental [f], pronounced with the lower lip touching the upper teeth; while Quenya f had been a bilabial [ϕ], with friction produced between the lips, at the time of the change of fu > hu.

(2) This marks a divergence from the former Latin-like change of *θ > f initially (cf. *θindo > findo, LR p. 392 THIN-). It also adds, for the first time, a completely new phoneme /θ/ to the Telerin sound-inventory.

(4) The Telerin word for “terror” is quoted (PME p. 363) as gāialā. This is certainly an error. For one thing, the Sindarin form provided is goeol, where the o’s should reflect a Quendian (and Telerin) long ā (*gājāl- > *gaujaul > *goiol > goeol). For another, final long vowels were shortened in Telerin, except in monosyllables or as the result of late coalescence of vowels (e.g. aba+a > abā “don’t”). The word should therefore be read with the macrons transposed: gāiāla. The Common Eldarin form would have been *gājālā, with the adjectival suffix -lā added directly to the noun *gājā “terror” (T. gāia, S. goe, Q. āya). In Quenya, with the reduction of non-initial long vowels, it would have become !āyala. The initial long vowel, though immediately preceding a stressed heavy syllable, would not be shortened due to a stress that always fell, to some extent, on initial syllables (cf. The Road Goes Ever On, p. 60: “The initial syllable usually retained some degree of stress.”)

Notessi ar Mapai

The last large segment of Telerin words appears in a connected set of essays primarily relating to Eldarin number-words. They are cited from Vinyar Tengwar #47 and #48.

These fall into the following categories:

  • The Telerin numbers themselves;
  • Words for the hand, for parts of the hand, and actions of the hand;
  • Words for various members of the family, used as ‘play-names’ for the fingers of the hand
  • Miscellaneous words arising in the course of other commentary.

The first group is: 1. er or min, 2. tat, 3. nelet, 4. canat, 5. lepen, 6. enec, 7. otos, 8. (toloθ), 9. neter, 10. pai or pain. 11. minipe, 12. iūnec or iūnece (also spelt ‘Quenya-style’ with initial y- and k for c), 13. nelpe, 14. campe (spelt kampe), 15. lepempe, 16. enempe, 17. otospe, 18. tolpe, 19. neterpe. (48 pp. 6-8, 21; 47 p. 42)

For the most part these show no unexpected developments. The most common is the omission of a final short (non-high) vowel, e.g. *tata > tat, *nelede > nelet, *kanata > canat, *lepene > lepen, *enece > enec, *otoso > otos, *netere > neter. All of these also fail to show the syncope exhibited by Quenya: 3. nelde, 4. kanta, 5. lempe, 6. enque (instead of expected **enke, with -kʷ- substituting for -k-) 7. otso, 9. nerte.

3. Is followed by neled- in parentheses. Since Eldarin numerals do not inflect, presumably this reflects its form in composition. The stem is in any case neled-, and nelet could reflect a devoicing of final -d > -t. However, such a sound law cannot be deduced from this one example; conceivably neled would have been the regular form, but was changed to nelet by analogy with the preceding and following numbers tat, canat, both ending in -t. Re-formation of a word which is commonly recited in a customary sequence (like numbers or the letters of the alphabet) is, while not frequent, also not unheard of. In English, for instance, the name of the letter J (“jay”) is due to re-formation on the basis of the following letter K (“kay”; historically [kā]). In most other languages using the Roman letters, J is called jot, je, ji (and similar forms) but not ja.

In Latin, final voiced stops were not devoiced, but final stops are so rare in any case that almost anything to be said about them is exceptional. There are, for instance, only four words ending in -b (all prepositions) and only one ending in -p; a substantial but restricted class of words ending in -c (mostly reflecting truncated forms of the clitics -ce, -que) and no words at all ending in -g. Latin -d and -t are in low contrast, though there are two minimally contrastive pairs (ad vs. at, quod vs. quot). A final -d occurs only in a small number of function words (pronouns, conjunctions, and prepositions); it had once been much more common, but was lost, within historical times, following a long vowel (e.g. ekstrād extrad > ekstrā extra). Final -t occurs very frequently as the last consonant of inflected verbs, but is otherwise restricted to function words (including adverbs), the noun caput and derivatives of the latter like occiput. In Classical Latin final -t and -d were probably pronounced the same by some speakers; thus we find words properly spelled with -d sometimes written with -t (e.g. aput, haut, set for correct apud, haud, sed), a variation that continues into mediæval Latin. Conversely, in the Middle Ages we sometimes find -d for -t (e.g. capud for caput). There is no Latin word, however, that ends in -t but has a stem in -d-.

7. Was preceded by many variants (otoc, otoco, occo, osco) but otos seems to have been the final decision. (47 p. 42). Occo and osco represent conflicting (and perhaps both transient) ideas about the development of syncopated *otokō > *otkō in Telerin; either direct assimilation *-tk- > -kk-, or epenthesis of an s (otkō > *otsko) followed by simplification of the cluster > osko. In Latin secondarily arising *-tk- assimilates > -kk-: *sitikos > *sitkos > *sikkos > sikkus (siccus) “dry” (the Latin development of Proto-Indo-European *-tk- is unclear; the only certain example is *Hṛtkos > ursus, but the details of its transformation are doubtful).

8. Was apparently problematic, perhaps because Tolkien could not decide whether “eight” would be !tolot (< *tolodo) or toloθ, which appears in other notes (48 p. 21).

10. The source of the -n in pain is uncertain. The Quendian forms were *kʷaja, which regularly produces Telerin pai; but the alternative form *kʷajam (Vinyar Tengwar 42 p. 24) would produce either **paiam or **paian. Tolkien made contradictory statements made about the fate of final -m in Telerin; the example of edulam showed -m retained, and we read that “Common Eldarin final -m survived as such in Telerin” (47 p. 11); but then again we read that “Common Eldarin final -m became -n” (47 p. 24) and that Telerin final -n arose from either -m or -n (42 p. 24). The latter two appear to represent the last known decision on the subject. Tolken indeed considered, then rejected the forms paya/payan (48, p. 21). But contraction of paian to pain is an unlikely development in Telerin, where surviving CVCVC forms (like gaiar) are quite stable when not in composition. More likely pain is built to pai itself as stem, with -n either as an archaizing plural marker, or simply suggested by the -n- in the ordinal paiania (see below).

The numbers from 11-19 all show an extension -pe < *-kʷē, with shorter forms of the numbers 1 and 3-9. *-kʷē is supposed to be from a root √kewe-, “new, fresh” (48 p. 7). Practically, however, it acts just like English -teen, a modification of ten, or the similar Latin -decim < decem, and in the same way may be considered reminiscent of the word pai “ten” < √kʷaj-. 16 enempe seems to have been remodelled after 15 lepempe (instead of *enekkʷē > **eneppe); tolpe lacks any indication of the extension in the root √tol-od-, in this resembling nelpe beside nelet (√nel-ed-) and campe (<*kaŋkʷē or later can+pe) beside canat (√kan-at-).

The “hand” words of Telerin are:

1. mapa (<*ma-kʷā, literally “hand-ful(l)”) “hand,” superseding “hand” < *maȝa, obsolete in its original sense in Telerin, but surviving in the preposition “by agency of,” with genitive, e.g. !mā Faianāro “by Feanor”; also in the adjectives forma “right-hand,” θarma “left-hand,” and the noun mālime “wrist” (“hand-joint”).

Tolkien had originally considered māga from a root √mag- for “hand” before replacing it with from √maȝ-. This shows that Tolkien still considered that medial -g- was maintained in Telerin. (48 pp. 6-7, 18-20).

2. camba “the cupped hand” (47 p. 22).

3. pār “hand closed around a tool” < *kʷār < *kʷāra. This notably shows that original long vowels were maintained in Telerin monosyllables; but when in the final syllable of a polysyllabic word they were shortened, as in Telperimpar (sc. !telpērin+pār) “Celebrimbor.” This is exactly as in Quenya. (47 p. 8)

4. donda “fist” (47 p. 23).

5. plata “flat of the hand.” This shows syncope of the first vowel in a CVCVC root (√palat-, in this case) which either did not occur, or was reversed in Quenya. By this means Telerin acquired a number of initial consonant clusters (stop + l or r) which did not occur in Quenya. (47 p. 8)

The fingers were leperi, singular leper. From thumb to pinky, they were 1. nāpa, 2. tassa, 3. nellepe, 4. nente, 5. nīce (spelt nīke). (48 p. 5)

1. Had gone through several transient stages of invention: lepet > tolmo > nāpo > nāpa. Its dual, nāpat — literally “two thumbs” — actually meant the thumb and forefinger, held together. This is the only evidence of a Telerin dual ending -t. From a root √nap- take hold of, pick up, grasp with fingers.

2. Means pointer, indicator, from a root √tas-.

3. Has the literal meaning “middle finger,” with nel- “middle,” an ancient variant of ened-, and lepe, a more primitive form of leper “finger.” Also has the ‘play-name’ tolle “sticker-up,” with the same root √tol- seen in Q. and S. tol “island.”

4. nente is from ente “moreover, what is more” (<*enetē) with prefixed n-, likely influenced by the neighboring finger-names in n-. The theory that it is from √net- with n-infixation is implausible; both Q. and S. forms of the name of this finger show reflexes of *enetē (Q. lepente, S. lebent), and the true root is √en-, “once again,” not √net-.

5. Literally means “little” (finger), containing one of several similar roots meaning “little,” in this case √nik-. It also has the “play-name” pince (spelt pinke), from √pik- small. (48 p. 6) Related to the hand and finger words were the verbs lepta- “pick up,” matta- “wield,” and palta- “stroke.” The last shows syncope of the second vowel in the root √palat-. (47 pp. 6, 9-10, 24-5)

The family words were (48 p. 6):
1. emme, diminutive emmece “mama,” used as a name for the forefinger
2. atta, diminutive attace “papa,” used for the thumb
3. nette, diminutive nettice, “sister,” used for the ring-finger. (also 47 pp. 14, 32-3, 38)
4. hanna, diminutive hannace, “brother,” used for the middle finger.
5. vinice or vince (both diminutive) “baby,” used for the pinky. Spelt winike, winke, ostensibly Quenya-like spellings that reinforce the idea that Telerin v spells [w]. From the root √win- young, hence literally something like “youngling.”

3. was itself a diminutive of nēθa “sister” from a root √netʰ-. But nēθa is also given as a Telerin adjective “gay, lively, girlish,” which may however have been superseded. (47 pp. 14, 32-3)

4. was likewise a diminutive of hāno “brother,” root √kʰan-, with the general Eldarin change of initial *kʰ > *x > h. (47 p. 14)

Innovations

1) *eu > iu (Q. eu): T. ciuta-, ciure, ciurān-, Q. keuta-, keure, keuran- “renew, renewed, new moon” (48 p. 7).

2) *kʷ > p (Q. qu): T. mapa, Q. maqua, T. pār, Q. quār; T. telperimpar, Q. t(y)elperinquar. (47 p. 8).

3) *kʰʲ > θ (Q. hy): T. θarma, Q. hyarma “left-hand” (47 p. 6).

4) *nd > d / #_ (Q. n): T. damme, Q. nanwe “low tide” (48 p. 26).

5) *nm > mm (Q. nw): T. damme, Q. nanwe “low tide.”

6) *ȝt > tt (Q. ht): T. matta-, Q. mahta- “(to) handle.”

7) *ȝt > ːt : T. lūta- “(to) bend” (47 p. 12).

(3) shows the change from *kʰʲ > θ. This doubtless passed through the stages *kʰʲ > *tʰʲ > *tʰ > θ, with the merger of the palatalized velars with coronals, exactly as in *kʲ > *tʲ > t, e.g. *kjelepē > tel(e)pe.

(6) and (7) are in apparent contradiction. But the particular development of *ȝ may be conditioned by the preceding vowel, or, more likely, lūta- is a later denominal verb based on the noun “bow,” formed after the very early loss of *ȝ in these roots — so that in terms of Common Eldarin, the root is actually √lū- and not √luȝ-.

Other changes also occurring in Quenya

1) *i > e /_# (Q. same): T. luine “blue,” imbe “gully,” mālime “wrist,” nette “sister,” diminutive suffix -ce < *luini, *imbi, *mālimi, *nettʰi, *-ki (47 pp. 6, 14, 48 pp. 6, 24).

2) *ttʰ > tt (Q. same): T. nette “sister” from root √netʰ- (47 p. 14).

3) *ȝ > : (Q. same): T. “bow,” Q. “curve,” from root √luȝ, T. “pool” from √loȝ (47 p. 12).

4) *m > n /_# (Q. same): stated, but no examples given (47 p. 24, 42 p. 24).

(1) If i > e is, as it appears, the normal change, then nimbi (with short final i) must be Proto-Telerin, and would manifest in Telerin as !nimbe “white.”

Conservative tendencies

1) *θ = θ / V_V (Q. s): T. nēθa, Q. nēsa “sister” (47 p. 14).

2) *b = b / V_V (Q. v): T. abar, Q. avar “one of the Avari” (47 pp. 13, 24).

3) *d = d / #_ (Q. l, or n when another n follows): T. duime, duine, duita-, Q. luime, nuine, luita- “flood, river, inundation” (47 pp. 24, 26, 30), T. donda, Q. nonda “fist” (47 p. 23).

4) *d = d / V_ (Q. r) : T. ened- (48 p. 25).

5) *g = g / #_ (Q. zero): T. gampa, Q. ampa “hook” (47 p. 20).

6) *kl = cl / V_V (Q lk): T. aclar, Q. alkar, S. aglar “glory” (47 p. 13).

7) *w = v (Q. v): T. cēva, Q. kēva “fresh, new” (48 p. 7).

8) Absence of syncope: T. ened-, Q. ende “middle” (48 p. 25).

9) Retention of original long vowel: T. ciurān-, Q. keuran-, S. cýron “new moon” (48 p. 7).

(1) is notable as a departure from the Latin-type developments, where *nētʰā would have given **nēθa > **nēða > **nēda. Under this new rule, however, the reflex of *bretʰelē would be not bredele but breθele.

(2) also appeared in the plural form abari in Quendi and Eldar.

(6) exemplifies a construction in which a stem CVCVC with the addition of a prothetic vowel and with syncope, reappears as VCCVC: thus √kalar- (T. calar “lamp”) > *aklara (T. aclar “glory”). The same construction is seen in the Telerin word aplat “prohibition” from √palat-. The form aclar seems to contradict the tendency to insert a svarabhakti u in a *kl sequence, seen in tancula, heculo; but conceivably both developments were possible, as seen in Latin saeclum beside saeculum.

(9) The Telerin form is given as a stem ciurān- from *keurāna. Since final short a was lost in Telerin, the nominative form must end in -n; and since a stem, rather than a nominative form is given, the nominative must show a different form, presumably ciuran, with shortening of the final long vowel just as in pār and telperimpar. The Telerin nominative would thus be ciuran, the genitive ciurāno; but in Q. the comparable forms would be keuran, keurano, since non-initial long vowels were shortened in Q. (except in morphologically transparent compounds or prefixed words, like elen-tāri, ū+nōtima, and before certain affixes, e.g. -va).

In addition to the words above, notes accompanying the “numbers” essays provide an account of the development of Eldarin diphthongs in Telerin (48 p. 7) which will be summarized further down.

Duini Gondonōreo

Of the 21 Telerin words appearing in the “rivers and beacon-hills” essay, eight are employed to clarify the history of a handful of Sindarin place-names: Nîn-in-Eilph, Glanduin, Gwathló and Ringló; the remainder are number-words, mostly ordinals, intended to help clarify the construction of the name Lefnui.

The first eight are alpa “swan,” glana, glanda, glania-, and glanna “edge, boundary, (to) limit, limited,” lanca “sudden end,” loga, logna “fenland, soaking wet” (Vinyar Tengwar #42, pp. 7, 8, 10).

The phonological data demonstrated by these words are:

Innovations

1) *kʷ > p (Q. qu): T. alpa, Q. alqua, S. alph.

2) *j > i / C_V (Q. y): T. glania-, Q. lanya- < *(g)lanja-.

Conservative tendencies

3) *gl = gl / #_ (Q. l): T. glana, S. glān; T. glanda, Q. landa, S. glann; T. glania-, Q. lanya-, S. gleina- (sic; probably for glenia- or gleinia-).

4) *g = g / V_ (Q. zero): T. loga, S. (Q. **loa not found in the sense of “fen”); T. logna, S. loen (Q. **lōna not found in the sense of “soaking wet”).

(3) Either shows the preservation of an initial Quendian *gl-, or, conversely, the differentiation of roots beginning in l- by the optional prefixion of g-, in which case it is actually a Telerin innovation. The root in question appears as both √glan- and √lan- in both Telerin and Sindarin, but in Quenya only as √lan-: T. lanca, Q. lanka, S. lanc.

(4) The contrafactual Quenya forms are, in this case, Tolkien’s own. Both existed in Quenya, but originating from different roots and with different meanings: loa “growth, year” < √law-, lōna “pool” < √lon-. It is possible that the spelling of logna disguises a sound-change. In Latin -gn-, though frequently arising from *-gn-, was actually pronounced [ŋn] with regressive nasalization of the stop. Since Telerin also shows regressive nasalization (as in *kʰabnā > hamna), and since its spelling is influenced by Latin spelling conventions, conceivably logna is actually pronounced [loŋna].

The quoted Telerin cardinals in this essay are tata “two,” canat “four,” and lepen “five.” The forms are all conservative, though tata may show the loss of an initial vowel from *atatā (cf. Q. atta and perhaps T. vante for *avante) and canat, lepen have lost final short vowels (Q. canta, lempe; S. canad, leben).

The Telerin ordinals are given in full from 1st to 10th as follows: 1st minya, 2nd tatya, 3rd nelya, 4th canatya, 5th lepenya, 6th enetya, 7th ototya, 8th tolodya, 9th neterya, 10th paianya. (Vinyar Tengwar #42, p. 25)

There is an obvious problem with every single word in the above list: all of them end in -ya, which we have consistently seen elsewhere in Telerin turned into -ia. I am inclined to believe that this is an oversight, influenced perhaps by a close focus on the Quenya forms (some of which ended in -ya, i.e. minya, tatya, nelya, lemenya). The possibility remains that Tolkien had, at least briefly, considered restoring [j] post-consonantally in Telerin; but probably we should read minia, tatia, nelia, canatia, lepenia, enetia, ototia, tolodia, neteria, paiania, and I will quote the forms as such below.

These are evidently the cardinals with the suffix -ya (-ia) tacked on. In a few cases we see a change due to consonant contact, namely:

1) *-kj- > -ti- / V_V: T. enetia

2) *-mj- > -ni- / V_V: T. paiania

(1) Enetia is based on the stem *enek-; we are thus seeing in *enekjā > *enetja > enetia an instance of a palatalized velar becoming a coronal, similar to that in *kʲelepe > telpe, except that in this instance a reflex of the original *j remains. In Comparative Tables, the medial reflex of *-kj- was -ci-; but this was precisely because at that time the palatalized velars merged with the plain velars, not with the coronals. Given the latter, -ti- is precisely what we’d expect, though it is unlike Latin, where medial *-kj- > -ki-.

(2) Paiania is from a stem *kʷajam, adding *-jā to *kʷaja- + the plural suffix -m. Possibly we are simply to understand *kʷajam > *kʷajan > *pajan > paian, with the ordinal form paiania (rather than *paiamia) due to adding the suffix to the cardinal. In this essay (and in contrast to the form edulam quoted above), Quendian final *m had become -n in Telerin. But it is also possible that paiania comes directly from *kʷajamjā, and this is the more likely in that it disagrees in form from the cardinal pai/pain. In Latin, Indo-European *-mj- > *-nj- > -ni-; *gʷemjō > *gʷenjō > *wenjō > weniō (venio) “I come,” and the same change is plausible for Telerin; together with *-kj- > -ti-, it suggests a strong preference for sequences of coronal+j at an earlier stage of Telerin.

Ototia does not show a systematic sound change. It is a replacement of expected *otosjā > **otoria. The stem is explicitly given as otos-, but ototia appears due to the remodelling influence of other ordinals, of which the neighboring enetia is the most salient. Although in Sindarin there is a cardinal odog “seven,” this cannot be used to postulate a stem otok- which could given *otokjā > ototia; this is because odog is itself remodelled (from odo) on the preceding cardinal eneg “six”! In this version of the history of the numbers, the stem *otok- never existed in Sindarin, much less in Telerin.

A small handful of other forms are found elsewhere. In a late essay of about 1968, the form andanē “long ago, once upon a time” (literally: anda+nē “long it was”). Another essay of about the same time informs us that Telerin used the pronominal element de for the 2nd person ‘polite’ forms, both singular and plural, and lacked le. (Vinyar Tengwar #49, pp. 31, 51) (Hence we may guess that the Telerin equivalent of hiruvalye was !hirubadie or !hirubāde.)

The somewhat unfortunate Telerin name Teleporno for Celeborn appears in Unfinished Tales, p. 266, as part of what seems to be a still unpublished linguistic essay. According to this account, the Telerin word for “silver” was telpe, but telep- in compounds. This is in accordance with the appearance of telpe in The Etymologies as well as The Shibboleth of Feanor (PME p. 356).

However, in a letter of 1972 (The Letters of J.R.R. Tolkien, pp. 425-6, #347) — the very last writing on Telerin available to us — the reverse is true. According to this letter, the Telerin word was telepe (telepi in the published Letters is a misreading of Tolkien’s handwriting) and telpe- only “occurred in words of length such as Telperion.” Thus we have Telporno given in the letter as Celeborn’s Telerin name, not Teleporno. Although contradicting previous citations of telpe, the form telepe is completely consistent with the noted tendency of Telerin to avoid syncope; if *ŋgolodō gives golodo, then it is completely reasonable for *kjelepē to yield telepe. Under this rule or tendency, golodo could exist comfortably alongside words like goldōrin and goldolambe.

Lāmapenta Lindarinua

Putting all of the above data together, it is possible to arrive at a fairly complete account of the historical phonology of Telerin. In many ways, it is much simpler and more straightforward than Quenya (much less Sindarin); and the resulting system is more compact and certainly easier to pronounce. On the whole it continues the structure of Quendian faithfully, and its development is mostly marked by a simplification and reduction of certain Quendian phonological complexities, with very few new complexities introduced.

Vowels

Short:

i < *i and < *j between a consonant and a vowel
e < *e and final *i; initial e+nasal < syllabic nasals
a < *a
o < *o
u < *u and < *w between a consonant and a vowel; also ul < syllabic *l̩

Short vowels could also be derived from long ones by atonic shortening, e.g. lindalambe < *lindālambē. This shortening affected most final vowels in words of more than one syllable. The only original short vowels to survive in final position were *-i and *-u, the first of which became -e. All other final vowels in words of more than one syllable were derived from long *-ā, *-ē, *-ī, *-ō, *-ū, and were almost always short. Telerin words ending in consonants often ended in a vowel in Quendian; but already in Common Eldarin short *-a, *-e, *-o were lost after liquids and nasals l, m, n, r (47 pp. 13, 24).

Long

ī < *ī.
ē < *ē, *ei, *ez (before d in *ezdē).
ā < *ā, *ǣ, *ǭ.
ō < *ō, *ou.
ū < *ū

Diphthongs

ai < *ai.
ui < *oi, *ui.
au < *au.
iu < *eu, *iu

These diphthongal developments, from Vinyar Tengwar #48, p. 7, are the same as those in Comparative Tables, except that there *ei, *ou > T. ai, au rather than ē, ō. Some other diphthongs apparently arise secondarily, from a combination of originally distinct vowels; e.g. oi in elloi < *edlō+ī.

Consonants

Telerin has 16 consonants, with the following initial correspondences:

p < *kʷ, *p.
b < *gʷ, *b, *ŋgʷ, *mb.
f < *kʷʰ, *pʰ.
m < *ŋʷ, *m

t < *kʲ, *t.
d < *gʲ, *d, *ŋgʲ, *nd.
θ < *kʰʲ, *tʰ.
n < *ŋʲ, *n

c < *k.
g < *g, *ŋg.
h < *kʰ, *h.

zero < *ŋ, *ȝ

s < *s, *sʲ.
r < *r.
l < *l.
v < *w.
y/i < *j

Initial consonant clusters *st-, *sp-, *sk-, *sw-, *dr-, *gr-, *gl- were retained.
New initial clusters br, cr, cl, dr, fl, pl, pr, tr could be formed by syncope.
*sn-, *sm- > n-, m-.
*sr- > fr- (or θr-?)
*sl- > fl-.

*kʲ, *kʰʲ, *gʲ, *ŋgʲ/*ŋʲ, *sʲ became *tʲ, *θʲ, *dʲ, *ndʲ, *sʲ; when initial, they were depalatalized (and denasalized), as shown above, but medially they were retained as *tj, *θj, *dj, *ndj, *sj until the change of *j > i / C_V, and became ti, θi, di, ndi, ri. Medial *mj (never initial) also > *nj > ni.

Intervocalic developments were usually the same as above, but:
r < *s as well as *r. s < *s when *r is separated by only one syllabic nucleus
i < *j (perhaps merely orthographic)

Medial consonant combinations arising from change:

cc < *tk (?)
ct < *gd

ll < *ld, *dl, *ln.
lp < *lkʷ

mb < *ŋʷ, *ŋgʷ.
mm < *nm.
mn < *bn.
mp < *ŋkʷ.
mpt < *mt

rp < *rkʷ

sc < *zg (and *tk?).
sp < *zb.
ss < *ȝs

tt < *ȝt, *ttʰ

Medial combinations unchanged: cl*, ct, gn, lc, ll, lg, lm, lp, lt, mb, mm, mp, nd, nn, nt, ntr, ŋc, ŋg, pl, pt, rm, rn, sc, sp, ss, st, tt.
Medial combinations formed secondarily: ln < *lVn, ld < *lVd

*Also cl > cul, as in tancula, heculo.

Pettasampane

The morphology of Telerin was evidently much like that of Quenya. In the verb, evidence is found for the aorist, the present tense with lengthened stem-vowel, past tense in -ne and augmented perfect tense; in all likelihood there was also a future in -uba.

The verbal endings can be given at least two possible forms, depending on whether Telerin more resembled Quenya or Sindarin in this respect:

More like Quenya:

1st person sg. petinie, melanie pl. incl. petingue, melangue pl. exc. petimme, melamme
2nd person fam.sg./pl. petic(e), melāce resp. sg./pl. petidie, meladie
3rd person sg. pete, mela pl. petinte/petir, melante/melar

More like Sindarin:

1st person sg. petin(e), melāne pl. incl. petince, melance pl. exc. petimme, melamme
2nd person fam.sg./pl. petitie, melatie resp. sg./pl. petide, melāde
3rd person sg. pete, mela pl. petinte/petir, melante/melar

The noun was inflected much like Quenya; based on the scanty evidence, paradigms can be reconstructed as follows:

Nouns in -a, -ō

Case Singular Dual Plural
Nom/Acc: ciria ciriat ciriai
Genitive: cirio ciriato ciriaio
Dat/All cirian(a) cirianna ciriain(a)
Locative ciriare/cirias ciriasse ciriaire/ciriais
Ablative cirialo ciriallo ciriailo
Instrumental cirianen ciriannen ciriainen

Nouns in -e and consonant stems:

Case Singular Dual Plural
Nom/Acc lasse lasset lassi
Genitive: lasseo lasseto lassio
Dat/All lassen(a) lassenna lassin(a)
Locative lassere/lasses lassesse lassire/lassis
Ablative lasselo lassello lassilo
Instrumental lassenen lassennen lassinen

The possessive suffixes were likely as follows: 1sg. ciriania, 1 incl. ciriangua, 1 excl. ciriamma; 2fam. ciriatia, 2resp. ciriadia (sg. & pl.); 3s. ciriaria; 3pl. cirianta. Words bearing these suffixes could be further inflected like any other noun in -a (e.g. cirianiai “my ships”).

Estel Lindaio

Telerin obviously has much less direct evidence for it than either Quenya or Sindarin. On the other hand, none of Tolkien’s languages has so much indirect evidence pointing to it, since it is illuminated, from both sides as it were, by Quenya and Sindarin. A very substantial Telerin vocabulary can be built up on the basis of those languages; and although it is of course impossible to be certain, the likelihood is high that any given word with cognates found in both Quenya and Sindarin would also be found in Telerin. For instance, there can be little doubt, despite the lack of textual evidence, that the Telerin word for friend must be !mello, stone is !gondo, the sun is !anar (anār-), a word is !petta, and “glorious” is !aclaripa (Q. meldo, ondo, anar, quetta, alcarinqua; S. mellon, gond, anor, peth, aglareb).

Telerin therefore has just as much potential to be ‘reconstructed’ into a functioning language as Quenya or Sindarin. In addition, it has a flavor that is quite different from either of them: less decorative perhaps, simpler, but also more down-to-earth, more recognizable, familiar, and normal. Tolkien once wrote something of the same sort, not directly about Telerin, but the sentiment crosses language barriers:

“…I prefer Latin. I feel like… an eminent philologist… who once wrote to me about some language recently discovered: ‘It is of a kind that you and I both feel to be normal, in a central human mode — it indeed resembles Latin.'” (The Letters of J.R.R. Tolkien, p. 419, #338.)