[ This text was originally sent by David Salo to Elfling; I found it in old archives and liked it very much. When I was preparing my web page I contacted David and asked him if I can place his letter on my page. He kindly agreed and I am very thankful for it. I found it as one of the most interesting secondary texts on Tolkien's languages and writing systems and it will give you good impression about Quenya Tengwar as well as about Quenya phonology. --  I added few notes for better understanding; you will need Dan Smith' fonts, namely Tengwar Quenya to see the transcriptions. ]

        In a Very Small Footnote to Appendix E of The Lord of the Rings, J.R.R. Tolkien wrote "In Quenya in which a was very frequent, its vowel sign was often omitted altogether. Thus for calma 'lamp' clm could be written." On first reading, one may think (as I did) that this seems somewhat far-fetched: wouldn't this omission of a's cause a hopeless confusion? And when Tolkien continues "This [clm] would naturally be read as calma, since cl was not in Quenya a possible initial combination, and m never occurred finally. A possible reading was calama, but no such word existed" the confusion only increases. Just how are you supposed to know what are possible initial combinations and what consonants can and cannot occur finally in Quenya?
        Actually, you can get a pretty good idea by studying Quenya vocabulary and nomenclature; and when you combine the characteristics of Quenya phonology with the peculiarities of its system of writing, it turns out that the confusion resulting from a-omission is minimal. In fact, even if Quenya were written with no vowels at all, it would almost always be possible to determine the syllable-structure of a word (i.e., where the vowels fall) from the written form! But since that's far from obvious at first glance, it is worth discussing the characteristics of both the phonology and the writing, and how they interact.
        The answer to the question "what are possible initial consonant combinations in Quenya?" is "none", or -- if you wish to treat the labio-velars of the quessetéma and the palatoalveolars of the tyelpetéma as combinations -- "only a velar consonant (e.g. k) combined with w, or an alveolar (e.g. t) combined with y". In practice this means only ty, ny, kw (qu) and ñw; one might also include hy, hw, hl, and hr, but from a phonetic point of view these are much better regarded as voiceless y, w, l, and r (i.e., related to y, w, l., and r as t is to d, or f to v). So the number of possible initial consonants or clusters is very limited: I count only 22. [ 1 ]
        The number of possible final consonants is even more limited: only t, r, s, l, and n occur finally and then they always follow a vowel, alone, never being double or part of a cluster. The number of possible medial consonant clusters is greater -- I count about 58 -- but still relatively limited. [ 2 ]
        The importance of consonant clusters and final consonants is this: the usual structure of a Quenya word is a series of open syllables: C[onsonant]V[owel]CVCVCV.... One can guess that where a consonant appears, a vowel will follow directly behind. If, in writing, no vowel is indicated, one can guess that the consonant is followed by a -- unless there is a good reason to suppose that the next consonant follows without a gap. It helps, then, to know that -tr- is not a possible sequence in Quenya: it would keep you from reading atarinke ("Little Father") as **Atrinke.
        Back to "calma". Without reference to possible combinations and positions of consonants, it looks as if without the vowels, you could divide it into all kinds of syllables: could it be calm? clam? clama? aclam? acalam? aclm? aclma?
        Quenya spelling comes to the rescue in this, and many other cases. For starters: whenever a word begins with a vowel, there must be what is called a 'short carrier' at the beginning of the word. Being an infallible sign of an initial vowel, it follows that if it "carries" no vowel marking or 'tehta', that it must therefore be pronounced a. So the Quenya word alcar "glory" must be written: Short Carrier-Lambe-Calma- Óre. [ 3 ]
        The Short Carrier also appears whenever a dissyllabic sequence of vowels appears. For instance, the word fea "soul" must be written Formen(e)-Short Carrier. If this second short carrier carries no tehta, it must be an a. [ 4 ]
        The other infallible sign of a vowel is the Long Carrier, which appears whenever a long vowel (á, é, í, ó, ú) is written, in whatever position. So Áre "sunlight" can be written Long Carrier-Rómen(e), where the Long Carrier, without tehta, must represent long á. [ 5 ]
        Equally infallible syllable markers are the tengwar yanta and úre. These are used to represent the final element in monosyllabic diphthongs: yanta is the -i in ai, oi, ui, while úre is the -u in au, eu, iu. So if yanta or úre are found alone, without tehta, they must be read as ai and au, respectively. [ 6 ]

        Quenya spelling also significantly reduces the number of consonant sequences that raise the question "Is there a vowel between these two consonants or not?" The most obvious way in which it does this is by writing the consonant sequence with a single tengwa. Such tengwar (or ones which could be so considered) are quesse (kw), ando (nd), umbar (mb)  anga (ñg), ungwe (ñgw), hwesta (hw), anto (nt), ampa (mp), anca (ñk), unque (ñqu), nwalme (ñw), arda (rd), alda (ld), esse (ss).
[ 7 ]
        A second way to show the inseparability of two consonants was by writing them, not with two separate tengwar, but by a tengwa with a modifying sign. For instance, a bar placed below or inside a tengwa indicated that it was doubled: this was done with tinco, calma, malta, númen, rómen and lambe to represent the sounds tt, cc, mm, nn, rr and ll. Two dots below the tengwa show combination with y: this combined with tinco, ando, anto, númen, malta, rómen, lambe and hyarmen represented the sounds ty, ndy, nty, ny, my, ry, ly and hy. A curling shape attached to the 'bow' (lúva) of a tengwa represented a following s: attached to tinco, parma, and calma they represented the sounds ts, ps, and cs (this sometimes written x in Romanization, as in Helcaraxe). [ 8 ]
        The sign halla (a single upright |) indicated voicelessness in a following r (rómen) or l (lambe); a vowel following it should never be suspected. [ 9 ]
        These helpful 'abbreviations' are built into the writing system. But there are other, less obvious features of the system which, while no less helpful, seem adventitious.
        For instance, the choice between the tengwar, rómen and óre to represent the sound r is based on position in a word. When preceding a vowel (as initially, or between two vowels) or a semivowel (like y or w), rómen was used; but when final, or preceding a full consonant, óre was used. This means that if rómen is followed by a consonant, a vowel must intervene -- and this vowel, in the no-a system of writing, must be a. So arani "kings", written Short Carrier-Rómen-Númen(i) must be read arani, not arni; the latter word, if it existed, would be written Short Carrier-Óre-Númen(i). So the sequences rt, rty, rc, rqu, rm, rn, and rs will always be distinguishable from -rat-, -raty-, -rac-, etc.; because in the former the first element must be óre, and in the second, the first element must be rómen. It can also be told that, whenever óre ends a word, the last sound in the word is -r; but where rómen ends a word, and no vowel is indicated, it must be read -ra. So hára "sits" must be written Hyarmen-Long Carrier- Rómen; but hárar "they sit" is Hyarmen-Long Carrier-Rómen-Óre. [ 10 ]
        Even more interesting is the use of two letters, súle (or thúle) and silme to represent the sound s. Because of the different histories of these two s-sounds, it is easy to distinguish between an s followed by a vowel and one followed by a consonant. Clusters containing s include st, sty, sc, squ on the one hand, and sw, rs on the other. (ps, ts, cs, and ss have already been dealt with). The first four are necessarily written with silme; the last two, with súle. Since the original s sound -- represented where it survives by silme -- was altered to -r- between vowels (or a vowel and a semivowel), wherever silme appears it must be either initial, final, or attached to another consonant. If it precedes a semivowel, like w, it must have originated from *th, and so be written with súle. Likewise, r was assimilate to a following s (so *neresâ > *nersa > nessa) so the sequence -rs- (as in Narsilion) represents an earlier *rth. The rule then is: if a silme is followed by a consonant, expect no vowel in between; if súle is followed by a consonant (not a semivowel like w), expect a following vowel. [ 11 ]
        There were several other clusters ending in -w: tw, nw, rw, lw. Presumably these could be written with the 'w-curl', an attached tehta, avoiding all confusion; but since w did not usually occur between vowels, having either become v or disappeared in these positions, there should not normally be any difficulty. [ 12 ]
        The most confusing set of written clusters, then, are those following lambe, since this tengwa has no distinctive shape before a consonant or between vowels. Nothing tells you whether Short Carrier-Lambe-Quesse represents alqua "swan" or *alaqua. The same goes for the sequences lt, lc, lm, lv, lf. Only a handful of clusters, also confusing to some degree, remains: mn (malta-númen, which could be read -mn- or -man- (e.g. lamani vs. lamni "animals"); pt (parma-tinco), which could be -pt- or -pat- (e.g. hyapati "shores" vs. *hyapti); and ht, hty, which could be read either -ht-, -hty- or -hat-, -haty-; although instances of the former are far more common. Possibly h should be represented by the tengwa hyarmen when preceding a consonant (as in -ht-), but by aha (harma) when between vowels. [ 13 ]
        On the whole, then, Quenya writing is a system in which writing-system and phonology reinforce each other, so that even when the most common vowel is omitted, very little confusion is engendered, and that only in a number of predictable instances. With some knowledge of the vocabulary, the chances for error are even more reduced.

[ notes ]

Here follow a few editorial notes to the text. It was me who wrote them and not David, except for the first two. They are taken from a private correspondence between me and David. I hope I transcribed the Tengwar correctly. I would like to thank to Angasúle who kindly helped me with this.


[ 1 ]
They are: f, h, y, c, l, m, n, ñ, p, r, s, t, v, w, ty, ny, cw (qu), ñw, hy, hw, hl, hr. This is the state of Exilic Quenya (i.e. the language of Noldor who returned in M-e from Valinor), Valinórean Quenya yet possessed th, but later it merged with s. In the Third Age Quenya the number would have been a bit smaller, as hl/l, ñ/n, v/w would have merged (and ñw would be nw).
[ back ]

[ 2 ]
They are:
nasalized consonants : nt, nd, nty, ndy, mp, mb, ñc, ñg, ñqu, ñgw
geminates : tt, cc, mm, nn, ll, ss (pp should be possible but was not found)
r + consonant : rt, rty, rc, rqu, rd, rm, rn, rs, ry, rw, rv (rp should be possible but was not found), rr occurs, but apparently only in compounds : Earráme, Palarran; or from vowel+sr: mirroanwi; rf occurs in the compound Arfanyarasse
l + consonant : lp, lt, lc, lqu, ld, lm, ly, lw, lv (lty should be possible but was not found); lf occurs in a compound (alfirin)
consonant + s : rs, ps, ts, cs (x)  (e.g. Narsil, Narsilion; only from rth)
consonant + c : sc
consonant + t : st, ht
consonant + ty : sty, hty
consonant + y : my, (ny), ry, ly, (ty)
consonant + w : nw, rw, lw, tw, sw
consonant + qu : squ
nasal compound : mn
[ back ]

[ 3 ]
"glory" alcar : `ja6 [ back ]

[ 4 ]
"soul" fea : eŠ` [ back ]

[ 5 ]
"sunlight" áre : ~7Š [ back ]

[ 6 ]
ai : lE, oi : lY, ui : lU; au : .E, eu : .R, iu : .T [ back ]

[ 7 ]
quesse (kw) : z, ando (nd) : 2, umbar (mb) : w, anga (ñg) : s, ungwe (ñgw) : x, hwesta (hw) : c, anto (nt) : 4, ampa (mp) : r, anca (ñk) : f, unque (ñkw) : v, nwalme (ñw) : b, arda (rd) : u, alda (ld) : m, esse (ss) : k [ back ]

[ 8 ]
tt : 1/, cc : a/, mm : t/, nn : 5/, rr : 7/, ll : j/
ty : 1Í, ndy : 2Í, nty : 4Í, ny : 5Í, my : tÍ, ry : 7Í, ly : jÍ, hy : 9Í
ts : 1_, ps : q+, cs : a+ [ back ]

[ 9 ]
hr : 7½, hl : j½ [ back ]

[ 10 ]
rómen : 7, óre : 6
arani : `75È
*arni : `65È
rt : 61, rty : 61Í, rc : 6a, rqu : 6z, rm : 6t, rn : 65, rs : 63
rat : 71, raty : 71Í, rac : 7a, raqu : 7z, ram : 7t, ran : 75, ras : 73 / 78*
hára : 9~7
hárar : 9~76
* : Sequence 68 is not possible. [ back ]

[ 11 ]
súle (thúle) : 3, silme : 8
st : 81, sty : 81Í, sc : 8a, squ : 8z
sw : 3n, rs : 63 [ back ]

[ 12 ]
tw : 1é, nw : 5é, rw 6é: , lw : jé [ back ]

[ 13 ]
"swan" alqua : `jz
lt/lat : j1, lc/lac : ja, lm/lam : jt, lv/lav : j1, lf/laf : je
mn/man : t5; pt/pat : q1
ht : 91, hty : 91Í
hat : d1, haty : d1Í [ back ]

<get out get outta here enough already>