THE EVOLUTION FROM PRIMITIVE ELVISH TO QUENYA

BY

HELGE KRE FAUSKANGER

EDITED WITH NOTES AND COMMENTS BY

VICENTE S. VELASCO

REVISED, EXPANDED AND UPDATED VERSION

ARDALAMBION Of the Tongues of Arda is a website dedicated to the study of the languages created by J.R.R. Tolkien. Webmaster: Helge Kre FAUSKANGER. Email: Helge.Fauskanger@nor.uib.no THE EVOLUTION FROM PRIMITIVE ELVISH TO QUENYA is originally published in the Ardalambion website as an RTF (Rich Text Format) file. This can be downloaded at http://www.uib.no/People/hnohf/ qevolution.rtf. This PDF (Adobe Portable Document Format) file is an expanded and updated version with further annotations of Mr. FAUSKANGER and Mr. Vicente S. VELASCO. Set in AGaramond, Times New Roman and SILDoulosIPA Font Summer Institute of Linguistics (http://www.sil.org) Copyright information: While this publication can be downloaded for free from the Ardalambion website, this cannot be posted anywhere else in the World Wide Web without the expressed permission of the principal author and editor. Printed versions can be made and freely distributed among private individuals and groups, and not for commercial use.

TABLE OF CONTENTS
Preface Abbreviations Used in this Article INTRODUCTION PRIMITIVE QUENDIAN COMMON ELDARIN THE LOSS OF MEDIAL (OR H ) AND PRE-RECORD QUENYA BEGINNING OF THE RETRACTION PERIOD THE QUENYA SYNCOPE THE SPIRANTS ARISE THE SHORTENING OF FINAL VOWELS IN POLYSYLLABIC WORDS DEVOICING OF MEDIAL COMBINATIONS NEW VOWELS CHANGES AMONG MEDIAL CONSONANT CLUSTERS OTHER VOWEL CHANGES (INCLUDING DIPHTHONGS) CHANGES PRODUCING VALINOREAN QUENYA ANTICIPATORY VOWELS DEVELOPING FOLLOWING + CONSONANT SECONDARY SYNCOPE CHANGES PRODUCING EXILIC QUENYA APPENDIX: SOME SPECIAL CASES page 5 5 7 11 16 18 23 25 26 28 30 31 32 34 38 42 44 48 50 57

Preface by the Editor

This article was originally written by Helge Kre Fauskanger as part of his Ardalambion website. I have taken the liberty to edit the original text, limiting myself largely by correcting some grammatical errors, reorganizing and tightening the text from redundancies, as well as changing some of his orthographical conventions (mostly because of the limitations of the Rich Text Format) so it would be as close to his original intentions as I deemed possible. I have confined my own comments, additions and annotations to endnotes. Helges own notes originally interspersed in the main text are moved to the footnotes. He was also kind enough to send me revisions and additional annotations to be included in the final version.

Abbreviations Used in this Article

(All page references refer to the standard hardbound or trade paperback editions unless otherwise indicated.)

Etym: The Etymologies, in The Lost Road and Other Writings, pp. 347-400 Letters: The Letters of J.R.R. Tolkien. References refer to page number, not letter number. LotR: The Lord of the Rings. Page references are from the standard one-volume edition first published in 1991; all subsequent re-set editions have the same pagination. LR: The Lost Road and Other Writings. The History of Middle-earth, vol. 5 LT1: The Book of Lost Tales, Part 1. The History of Middle-earth, vol. 1 LT2: The Book of Lost Tales, Part 2. The History of Middle-earth, vol. 2 MR: Morgoths Ring. The History of Middle-earth, vol. 10 PM: The Peoples of Middle-earth. The History of Middle-earth, vol. 12 QL: Qenya Lexicon (Qenyaquetsa), published in Parma Eldalamberon 12 RGEO: The Road Goes Ever On. SD: Sauron Defeated. The History of Middle-earth, vol. 9 Silm: The Silmarillion. As published by Houghton Mifflin: both the First Edition (1977) and the Second Edition (2001) have identical paginations. WJ: The War of the Jewels. The History of Middle-earth, vol. 11 UT: Unfinished Tales of Nmenor and Middle-earth. VT: Vinyar Tengwar. Journal of the Elvish Linguistic Fellowship and the Mythopoeic Society. References refer to issue number and page. CE: Kh: PQ: Q: S: T: Common Eldarin Khuzdul Primitive Quendian Quenya Sindarin Telerin

INTRODUCTION
Whereas many students of J.R.R. Tolkien's languages would be inclined to focus on their classical formthat is, Quenya and Sindarin as they are supposed to have existed at the time when the narratives of The Lord of the Rings and The Silmarillion took placethe author himself is often seen to focus rather on the long evolution of these languages from the earliest forms of Elvish. Observes Christopher Tolkien, My father was perhaps more interested in the processes of change than he was in displaying the structure and use of the languages at any given time (LR:342). Elsewhere, he comments on how his father worked out a minutely refined historical development of Quenya and Sindarin (PM:367). To the academic, the fact that the historical development was so important to Tolkien should be sufficient to warrant a careful study of it. A deeper understanding of how the languages were supposed to have evolved during the long ages of Arda will also lead to a fuller apprehension of the method, spirit and focus of Tolkien's work. Yet this study does not have to be wholly academic: People who want to develop a Tolkien-language like Quenya into a useable system, even coining new words from Tolkien's roots to supplement the existing vocabulary, must know what rules Tolkien himself applied when developing Quenya vocabulary. It is a matter of debate whether a post-Tolkien Quenya word can be counted as a genuine word, but everybody would agree that a word that was not derived in accordance with Tolkien's rules and does not fit into his system couldnt be counted as genuine in any sense. What follows is a comprehensive (not exhaustive!) list of phonological changes that occurred during the evolution of Quenya. Though I briefly describe the nature of the changes, I have focused more on providing a survey of the observed changes than on formulating abstract rules that may seem to govern the sound-shifts. The changes have been listed in what may be assumed to be something roughly similar to their chronological order. The material has been grouped under the following headings: Primitive Quendian Common Eldarin Pre-Record Quenya Changes Producing Valinorean Quenya; and Changes Producing Exilic Quenya. In some cases, it is obvious where a certain phonological change belongs (Tolkien may even have made an explicit statement about it). In many cases, it is less than obvious. It should be understood that this is in no way a canonical of definite listing as regards the chronology of Quenya sound-shifts. This said, the material here presented is quite sufficient to explain how the vast majority of Quenya words reached their classical or contemporary form. I mention many of the clues and deductions that may throw some light on the chronological order of the changes, but when one studies these things in detail, it soon becomes obvious that the material is not entirely consistent. For instance, Tolkien seems to have experienced considerable trouble trying to figure out precisely when the shortening of

8
The Evolution from Primitive Elvish to Quenya the long final vowels of Primitive Elvish took place. Some of his reconstructed forms may seem to imply that this happened already during the Common Eldarin stage. On the other hand, the form Valinr Valinor in WJ:413 implies that the long final vowels were still in place as late as after initial b- had become v-, a change entirely characteristic of Quenya. (See also Appendix: Some Special Cases.)

Some practical matters: In his notes, Tolkien mentions many primitive, reconstructed forms, and usually he even places an asterisk before such ancestral forms to mark them as unattested. Since Tolkien's own supposed reconstructions must be assumed to be rather more authoritative than my own, much more genuine reconstructions, I have distinguished between them here. Tolkien's alleged reconstructions are marked with the symbol instead of *, the asterisk being reserved for my own reconstructions forms that are really unattested in Tolkien's published material. (A double asterisk ** marks a wrong form.) Most of the forms here marked are found in Etym. To the extent primitive forms here mentioned cannot be readily located, the wordlist appended to my article about Primitive Elvish will provide references to the primary sources.1 In the sources, Tolkien usually marks long vowels by means of a macron, a line above the vowel, so we will use them here as well: , , , , . However, in the Quenya forms we will eventually reach, long vowels are marked by means of an accent instead: , , , , . The nasal sound ng as in English king is here represented by , a spelling often used by Tolkien himself2. However, in accordance with Tolkien's own spelling the clusters ng and nk (nc) in the middle of words are so represented, though this is technically g, k. Syllabic consonants are here separated from the preceding part of the word by means of a dot, e.g. *hekl (the final l being syllabic, somewhat like English little is pronounced litl.) An initial syllabic consonant is followed by a dot: *ndn. A syllabic consonant in the middle of a word is both preceded and followed by a dot: kwentro. In Tolkien's spelling of primitive forms, the semi-vowel y as in English you is inconsistently represented as either y or j (and in Etym as published in LR, Christopher Tolkien in most cases altered his father's original spelling, changing j to y). In this study, we have consistently used the grapheme j when actual, primitive words are quoted, switching to y when we reach Quenya forms. (However, for easy reference, we maintain y when referring to entry-heads in Etym, e.g. YUR rather than JUR. Inevitably there are here some orthographic inconsistencies, as when we quote the primitive form wj while referring to the entry WAY in Etym, but the graphemes j and y in any case refer to the same sound in all Elvish forms.) The aspirated sounds of primitive Elvish are here marked by means of a raised h, as in ph, th, kh. (In published sources, Tolkien simply wrote ph, th, kh.) Elvish has certain labialized sounds, consonants pronounced with poised lips. The labialized quality of the consonant is here indicated by means of a raised w, e.g. kw, gw, nw, w. (However, in normal Quenya orthography kw is written qu3, and the other groups are represented as simple digraphs in -w, e.g. nw. The distinction between nw as a unitary labialized consonant nw and nw as a genuine cluster n + w is not upheld in the normal orthography. Indeed Tolkien does not uphold this distinction in any of his published writings.)

INTRODUCTION

Palatalized consonants are here indicated by means of a raised j, e.g. nj (= palatalized n, like Spanish ) and tj (more or less like the initial t of English tune). In Quenya orthography, the same sounds are represented as ny and ty, though these combinations may also represent genuine clusters n + y, t + y. The English alphabet has separate letters for only two spirant sounds, f and v. We will also use the special letters (more or less = English th as in think) and (= voiced English th, as in these). The back spirant, spirant g, Tolkien sometimes represented as , and we will use here as well. For ach-Laut in the ulterior forms we will use Greek letter (as ch in Scottish lochnot as in English church, a sound that does not occur in any known Eldarin language), but we will represent the same sound as ch in Sindarin words. Unvoiced w, n, m are here represented as digraphs hw, hn, hm. The digraph hy likewise represents voiceless y (= German ich-Laut), as in the normal orthography of Quenya. It will be noted that by introducing these special spellings, the exact pronunciation of primitive words reconstructed by Tolkien is made more specific than his original spellings are. While I do believe the interpretations here set forth are generally sound, they should not be taken as Tolkien fact. For instance, the word here cited as kwentro (with the symbol to indicate a primitive form actually given in Tolkien's material) simply appears as kwentro in the source. While there are good reasons to assume that the initial kw is meant to be a unitary, labialized consonant kw, and that the r at some point became syllabic, these phonological details are not directly confirmed by Tolkien's spelling and do not have absolute authority. As this discussion reaches a recognizable form of Quenya, we introduce LotR-style Quenya orthography. (The relevant spelling conventions are set out in the Introduction to my Quenya course.) We have already commented on the use of y instead of j; also notice the use of c rather than k as well as x rather than cs/ks, plus the marking of certain vowels with diaereses ( is so marked when occurring finally and in the combinations a, o). These divergences from the spelling of primitive forms do not suggest any phonological distinctions. We will now attempt to survey the changes that occurred during the evolution of Quenya from the earliest forms of Elvish, as far as Tolkien's intentions can be reconstructed. It must be understood that this discussion focuses on phonological developments: Of course there were also changes that had nothing to do with the regular sound-shifts (as when the old plural ending -i was in many cases replaced by the ending -r according to PM:402 a new device).

EDITORIAL NOTES
1 2 3 This can be downloaded at http://www.uib.no/People/hnohf/primelv.htm. Although he also used special forms of the letter n to express the same phoneme, like as in oldo (See in particular the pre-1994 Ballantine paperback edition of The Return of the King, p. 500). I believe this would be a good place to insert a differing opinion made by Lukas Novak in Elfling message 7706 (January 2, 2002). He writes:

10
The Evolution from Primitive Elvish to Quenya
Throughout your [Helges] article, you maintain that the [kw] and [kj] graphemes can sometimes mean unitary labialised/palatized consonants, since inter alia a) they are allowed to stand initially; and b) they allow syncope virtually producing a [three]-consonant cluster [but see p. 26, footnoteEd.]. This does not seem to me so convincing. The main reason is that you have to assume then that there are both genuine clusters of this type and their unitarized counterparts in the language. That seems to me very probable (no explicit data confirm this; Tolkien never [mentioned]any difference in [the] pronunciation of these letters, although he [indulged] himself on such minor variations of pronunciation like nasal n vs. [?], lw vs. lv > lb, slight palatization of of l in el, il, etc.) I think the data could perhaps be better explained by [simply] acknowledging that these clusters have [a] somewhat exceptional position in the languagedue to the [semi-vocalic] character of the y and w sounds. They simply share some characteristic of the clusters (most markedly they [lengthen the syllable in which they occur]), but not all of them [Emphasis on the original]. Your argument that they are unitary consonants because all clusters share certain characteristics, but these supposed clusters do not share all of them is circular, since you must presuppose that they are not clusters in order to have the major premise true. If I am right, then the sound hy should be considered only as an ich-Laut [], but as a similar semi-cluster ich-Laut plus y [j]. Otherwise there would have to be two phonemes in Q: si[m]ple ich-Laut in initial position (e.g. Q hyarmen) and an ich-Laut plus y medially (or only in accented syllable?), which seems also improbable. My thanks to Mr. Novak for letting me quote him directly for this revised version of the article, which I edited for clarity.

PRIMITIVE QUENDIAN
The primeval stage of Elvish speech, the language originally invented by the Elves at Cuivinen, cannot be seen as a static and unchanging entity even within the primitive period. Tolkien explicitly refers to the common development of primitive Quendian (VT39:8, emphasis added). There would be divergent developments even within the most primitive period, but there was apparently one unique point in pre-history when the various ultra-primitive language-constructions were unified and coordinated so that all Elves achieved one single language: The hundred and forty-four Quendi [that originally awoke] dwelt long together by the lake, until they all became of one mind and speech, and were glad (WJ:423). Tolkien wrote in an early source: Roots...are not words in use at all, but serve as an elucidation of the words grouped together and a connection between them (LT1:246). In all later forms of Elvish, the roots (stems, bases) can certainly be treated as abstract raw material for actual words, mere skeletons to be fleshed out with various derivational endings, producing the kind of words that can actually be used in speech. Yet if there was ever a stage where the roots themselves existed as actual words, it must be the most primitive form of Elvish. At the very beginning, we must probably imagine a fleeting era when as yet undifferentiated morphemes exist in a linguistic limbo, predating all derivation, all inflection, all distinguishing between different parts of speech. The first word ever spoken by the Elves was supposedly ele, a primitive exclamation, lo! behold! made by the Elves when they first saw the stars (WJ:360). Later proto-words, if that is what they are, seem in most cases to have consisted of two (usually identical) short vowels separated by a medial consonant. Usually there was also an initial consonant, though it could be missing (as in ele). Forms later considered roots, such as DELE walk, HEKE aside or ABA refuse (WJ:360,361), may be seen as representing actual, ultra-primitive words. This pattern came to be especially preferred in the case of stems with a verbal significance. At some point there occurred a reorganization of [...the] basic structure of the language (WJ:392), some originally monosyllabic stems being expanded to conform with it. Hence the primeval root having to do with vocal speech, KWE, which also appeared in elaborated forms KWENE and KWETE (the former underlying the very word Quenya, the latter producing the Q verb quet- say, speak). The reorganization Tolkien hints at seems to suggest that the first Elves eventually developed a feeling for distinct parts of speech, their language gradually assuming a more well-defined structure. There appear derivational elements assisting the transfer of morphemes from one part of speech to another, or modifying the meaning of the proto-words, which eventually begin to assume the function of roots from which actual words are derived. Later, in Valinor, the Eldar would analyze a Quenya word like tuil spring as representing tu-yu-le, becoming tuyle > tuile by the loss of the second vowel in the root TUJU (TUYU) sprout, bud (VT39:7). Tolkien seems to hint that this analysis was essentially correct, or at least later Loremasters are said to have held similar viewswhen Quendian origins had been further investigated (VT39:11 note 6). Given the form of the root TUJU, it might seem that the word for spring was originally simply this verbal root with the abstract ending -l attached: hence *tujul sprouting, budding.

12
The Evolution from Primitive Elvish to Quenya However, Tolkien in Etym listed tuil as the primitive form producing Quenya tuil. One of the very first phonological changes in Quendian may have been a primitive syncope whereby the second of two identical short vowels in many cases disappeared before a derivational ending:
*tujul > (*tujl >) tuil spring (or budding)

Many of the words that seem to have been syncopated very early involve two-syllable bases (sometimes original, sometimes extended from simpler roots, e.g. NAK from NAK or WENED from WEN[E]). In most cases only the fact that Tolkien listed the base itself (with the second vowel intact) allows us to deduce or theorize that some kind of syncope is operating. Some examples:*
From the base SAT: *asat > ast dust (Quenya asto, Sindarin ast) From NAK: *anak > ank jaw (Q anca, S anc) From PARAK: *parak > *park dry (Q parca, S parch) From TWAR: *tawar > taur forest (Q taur, S taur) From TLEK: *telek > *telk leg (Q telco, S telch)1 From LUG: *ulugund > ulgund monster (Q ulundo, S ulun(d)) From WENED: *wened > *wend maiden (Q vend, S gwend)

Here we list the eventual Sindarin forms as well as the Quenya forms, in order to demonstrate that these vowels must have been syncopated quite early. In Quenya, the second of two identical short vowels would have a second chance to become syncopated, but this happened after the branch of Elvish leading to Sindarin had separated from the branch leading to Quenya: When Quenya and Sindarin agree in having syncopated a certain vowel, we can often assume that the syncope occurred already in their common ancestral language. In Etym, the entry USUK provides a good example: Tolkien listed a primitive word us(u)k-w reek, producing Q usqu and S osp. It seems that ultra-primitive usuk-w turned into usk-w (or *uskw, the labio-velar kw becoming p in the Lindarin branch that Sindarin is a part of). *Usukw with no syncope might (via Old Sindarin **usupe, **uhupe) have produced Sindarin **uhub instead of osp. In the branch of Elvish that leads to Sindarin, an unaccented vowel in the first syllable of a word is very often syncopated, resulting in an initial consonant cluster: for instance, the base BORN- produces Sindarin words in br- (e.g. brn old from b'rn, the first vowel of the original base dropping out). However, in the branch of Elvish that leads to Quenya, the first vowel of a base is nearly always preserved even where it was originally unaccented. We seem to have only one divergent example: the Quenya word rca wolf Tolkien referred to d'rk, a syncopated form of the stem DARK. Actually the complete primitive word must have been *d'rk, reduced from even earlier *dark. As is evident from the example *tujul > (*tujl >) tuil spring, a semi-vowel j occurring at the end of a root might turn into a full vowel i when it occurred in front of a consonant (or was brought into contact with one after the syncope). In a similar environment, w became u. In Etym, the bases TWAR (extended from TAWA, VT39:7) and TUY (= TUJU, VT39:7) produce primitive forms like taur forest and tuil spring. Such words may be reworked from *tawr and *tujl , the syncopated forms of *tawar and
*

It will be noticed that Tolkien in many cases indicated (by means of an accent) that it is the first vowel of the relevant two-syllable bases that receives the stress. Where the stress was on the second vowel, no syncope occurred at this stage. Thus mork bear from MORK did not become **mork (though the word did eventually end up as morco in Quenya, because of a later syncope).

PRIMITIVE QUENDIAN

13

*tujul . Cf. also one word derived from the base LAWAR: laur golden light (for *lawr?) The syncope of the second of two identical short vowels in concomitant syllables is not entirely regular in the primitive language. There are quite a few examples of words that escaped this syncope. Examples of words in which the second vowel was not lost include galad tree, khjeles2 glass and kjelep silver (cf. Sindarin galadh, hele, celeb with the second vowel still intact: though it was eventually syncopated in Quenya al'da, hyel'l, tyel'p, this is a later development). It may indeed be doubted whether there was a primitive syncope at all, since some words mysteriously seem to have escaped it. When Tolkien lists a word like hekla outcast as a derivative of a root he quotes as HEKE (WJ:361) are we to assume that hekla must necessarily be shortened from even earlier *hekela by a primitive syncope, or did the derivational patterns of Primitive Quendian allow the suppression of the second vowel of the root from the start? Regarding the situation in Common Eldarin, Tolkien did note that both leaving out and retaining the second root-vowel in derivatives were legitimate and regular procedures (VT41:9) apparently simply a matter of choice. Anyhow: whether in PQ we have a somewhat sporadic and unpredictable syncope, or a somewhat unpredictable suppression of reduplicated root-vowels in derivatives, is mainly an academic question. Of course, Tolkien may have imagined it differently at different times and in any case even he himself preferred to deal with Primitive Quendian as a somewhat vague and hypothetical entity: The precise details were not necessarily to be pinned down in all cases, but were allowed to remain obscured by the passage of long ages. Besides this early syncope, if it ever occurred, some final vowels may also have disappeared in the early periodso early that their former presence seems to have no influence on any of the later stages. In VT39:6, Tolkien states that in Quenya, all final consonants had probably lost a vowel, if remote Quendian origins were considered. Earlier, in Etym, he had nonetheless cited a few forms that he himself identified as Primitive Quendian even though they ended in a consonant: atar father, dr man3, m hand.4 Perhaps we are to understand that there was an even more primitive stage of PQ where these words did have a final vowel, but what its quality might have been, we can only speculate about. The normal plural marker - evidently had to be shortened following a long vowel (or any vowel?); the Primitive Quendian plural Lindar (Elves of the Third Clan) is given as Lindi, not **Lind (WJ:378). If the latter form ever existed, it must have been altered very early. Perhaps the plural of a word with a short final vowel, such as swanda sponge, likewise appeared as *swandai rather than *swanda (the ultra-primitive form?) But a short final -e seems to have been displaced before , as when the plural of kwende elf was kwend (for ultra-primitive *kwende?) We will list some more very early changes, but whether they occurred in late Primitive Quendian or early Common Eldarin is difficult to say. Tolkien stated that the metathesis bm > mb occurred in Common Eldarin at the latest, but added that it was possibly earlier (WJ:416), which necessarily implies that it may have occurred already in Primitive Quendian:
labm > lamb [physical] tongue (Q lamba) labm > lamb tongue, language (Q lamb)

14
The Evolution from Primitive Elvish to Quenya (Given the form of the root LABA whence these words are derived, we may perhaps assume that their ultra-primitive, pre-syncope forms were *labam, *labam.) The change bm > mb is an example of metathesis of a stop and a nasal sharing the same place of articulation; this is also the case with dn > nd, which seems to be another example of very early metathesis:
adn > *and gate (Q ando) sjadn > sjand cleaver (Q hyando)

A long vowel -, -, -, or - + -i becomes a diphthong -ai, -ei, -oi, or -ui:

Lindi > Lindai Lindar (WJ:378 vs. WJ:385), sg. Lind *orni > ornei pl. trees (SD:302), sg. orn

(If the combination -i ever occurred [e.g. *tri queens?], it was presumably contracted to -.) The combination Tolkien spells kw is probably to be taken as a labialized sound kw; thus, this is a single consonant rather than a cluster k + w. However, it seems that even genuine clusters k + w very early were assimilated to unitary consonants kw:
hek-w (root HEK(E) + ending -w) > *hekw (Q hequa except) salk-w (root SALK + ending -w ) > *salkw (Q salqu grass)

Notice that the latter example will later undergo syncope in Quenya: the second a has disappeared in the Quenya form. However, as will be discussed below, such syncope did not occur in front of a consonant cluster. We are therefore able to tell that by the time the syncope occurred, the original cluster kw had merged into a single labialized consonant kw. This kw survived into Quenya (in regularized spelling written qu, but the fact that Tolkien in early sources represents it by means of the single grapheme q also suggests that it is to be taken as a single consonantsalqu is indeed spelt salqe in the primary source). The aspirates are reduced to unaspirated stops before certain consonants. The aspirates kh and th (we have no example for ph) evidently become simple k, t before s:
*lokhs (root LOKH ) > *loks (Q lox = loks) *khoths (root KHOTH ) > khots assembly (no known Q descendant)*

Possibly, kh becomes simple k before t: The Quenya verb rihta- to jerk may descend from either *rikt or rikht; Tolkien gave the root as RIK(H), either RIKH or RIK. But the Quenya form would be rihta- in any case, so we cannot be sure.

EDITORIAL NOTES
1 2 In an earlier draft of this treatise, the ulterior forms of telco were assumed to be *teleku, *telku. However, Helge pointed out the plural form telqui is said to be analogical, evidently suggesting that it is not really justified by an archaic form *telku rather than *telk. However, in a later source, Tolkien changed the etymology, with S. heledh a loan-word from Kh. kheled (Silm:360, cf. the place name Kheled-zram, the Mirrormere), apparently obsolescing the root

The example khots, reconstructed by Tolkien himself, could also be taken to mean that aspirates could not occur before s even in Primitive Quendian, some phonological restraint prohibiting such combinations. The question has only academic interest.

PRIMITIVE QUENDIAN
KHJEL(ES)

15

3 4

and its derivatives in Etym (LR:365). Helge wrote that he included this example for purposes of analyzing phonological development in Quenya, as an attested example of how the primitive phoneme *khj- developed in Quenya. The only other example of *khj- that has by all indication survived externally is the stem KHJAR. While no attested ulterior forms are attested as of yet, the Q form hyarmen south can be reasonably speculated to have descended from *khjarmen. This example can perhaps be seen as a variant form of *nr < NERE. But in both earlier and later sources (LT2:339 or QL:57, and VT39:11), the primitive word for hand is said to have been maha.

COMMON ELDARIN
Common Eldarin was the language spoken by the Eldar (as opposed to the Avari) during the Great March from Cuivinen. This is the last common ancestor of all the later Eldarin tonguesQuenya, Telerin, Sindarin and Nandorin. Common Eldarin should not be thought of as an entirely homogeneous structure; apart from the linguistic changes that occurred during the March, there were already some minor dialectal differences. For instance, it seems that the Teleri had already turned the original labialized sound kw into p (whereas it remained kw = qu among the Noldor and the Vanyar). The aspirate th and the voiced stop d both become s before t:
*khotht > *khost to collect (Q *osta- > hosta-; cf. root KHOTH) *kotht > *kost to quarrel (Q costa-; Tolkien changed the root from KOT to KOTH) *ledt > *lest to leave (Q lesta-, cf. the entries LED, LED in Etym)*

Another example of dt > st was wed-t > *west to swear (Q *westa > vesta), but Tolkien dropped this word because it clashed with [*best >] vesta to marry. See entries WED and BES in Etym, and cf. lesta above. A t may also become s before another t when they are perceived to belong to different morphemes:
*put-t (root PUT + verbal ending -t) > *pust to stop (Q pusta) *lot-t (root LOT + the same ending) > lost to bloom (Q losta)

The latter example comes from VT42:19, Tolkien explicitly commenting on the change t-t in inflexion > st. This change does not occur when tt occurs within a unitary word. Indeed the same root PUT (see LR:382 s.v. PUS) that is the source of pusta- also yields Quenya putta stop (in punctuation), which is clearly meant to represent primitive *putt. It may be similar in sound to the primitive word that produced Q pusta, but here the double tt occurs as a medial fortification of the final consonant of the stem PUT. This *putt is therefore perceived as a unitary word, and the change tt > st does not occur. The above-mentioned Quenya word costa to quarrel was originally an example of tt > st, Tolkien referring it to kot-t[], but then he changed the root from KOT to KOTH. The nasals m and both become n when preceding a t or d:1
*kemtan > *kentan potter (Q centano the first element seems to represent the root KEM earth) *pilimda (final vowel uncertain, may also be -o or -e) > *pilinda arrow (Q pilin with stem pilind-; cf. the base PLIM, LR:382) *wit > *wint to scatter (Q winta > *vinta; if this word is to be related to the noun wing as suggested in PM:376, it seems to require a root *WI)

Final -w changes to w:
nidw > *nidw cushion (Q nirwa)

Final -m becomes -m after a dental consonant:

jatm > *jatm bridge (Q yanw)

It is not entirely clear whether the word lesta was rejected or not; the entry LED where it occurred was replaced by another versionbut the phonological rule dt > st may be valid all the same.

COMMON ELDARIN

17

(Cf. also Old Noldorin/Old Sindarin katwe from earlier katw, demonstrating that -w likewise becomes -w > -we following a dental consonant see the entry KAT in Etym.)
Medial w turns into gw:
liwi > *ligwi fish (Q lingw) newi > *negwi nose (Q nengw)

Evidently quite early (or even from the start), a long final vowel is shortened if the word in which it occurs appears as the first element of a compound:
and long + -mbund snouted = andambnd (not *andmbund) long-snouted lass leaf + -kwelne fading = lassekwelne (not *lasskwelne) autumn tuil spring + lind singer = tuilelind (not *tuillind) spring-singer, swallow

In some cases, but not regularly, the nasal element of the initial nasalized stops nd, g, mb develops into an independent syllabic consonant:
ndn > *ndn sunset (Q andn) *gj > *gj grandchild, descendant (Q indyo) *gwal > *gwal torment (Q ungwal) *mbarta > *mbarta fate (Q umbar)

In origin, this development seems to have functioned as a kind of strengthening that may also modify the meaning of the word: From gl philosophy is derived *gl deep lore, magic, and these words continue to coexist (eventually producing Quenya nl and ingol, respectively). But in most cases the original form with no syllabic nasal seems to have been lost completely, replaced by the strengthened form. Later Quenya developments suggest that the initial nasal of the group gw- became labialized by assimilation to the following gw, so that the whole group is technically wgw (or wgw where the nasal came to be syllabic); perhaps *gwal above should more strictly be represented as *wgwal. Some cases of haplology:
kukw > *kw or *kuw dove (Q cua) tuilelind > *tuilind spring-singer = swallow (Q tuilindo)

The changes above may have occurred in the earliest stages of Common Eldarin. Some of the changes listed below may be imagined to have occurred later say, after the Eldar had crossed the Hithaeglir. (By this time, Common Eldarin would not be entirely common anymore, since the Nandor had already left the march.) It may seem that relatively early, the aspirates th, ph, kh became normal t, p, k immediately following another consonant. The Quenya word ilfirin immortal is evidently meant to be derived from *lphirin- (root PHIR), but in the relevant entry in Etym, Tolkien seems to indicate that this should regularly have produced Quenya *ilpirin instead. The actual form Quenya ilfirin was apparently reformed after firin mortal (< *phirin-). During the Common Eldarin stage, tkh was apparently assimilated to kk. Perhaps tkh first became tk in accordance with the rule set out above, and this in turn becomes kk (evidence from Sindarin indicates that tk of whatever origin normally became kk, whence Grey-elven ch). The later Quenya word eccaira remote, far must come from *ekkair, but since the root is given as KHAYA, we must probably assume that at an even older stage, the word appeared as *etkhair with the prefix et- out- prefixed.

18
The Evolution from Primitive Elvish to Quenya In one attested case, tk does not become kk, but undergoes metathesis to kt instead: et-kel *out-flow early becomes *ektel (later ektele, whence Quenya ehtel spring, issue of water). S is voiced to z before a voiced stop:
esd > ezd rest (WJ:403 seems to place this change in the CE era)

Another example of the same would seem to be mazg pliant; given the base MASAG, this must come from earlier *masg. D becomes t before s:
sjads > sjats cleft, gash (Q hyats)

Before an unvoiced consonant, the back spirant lost its voicing by assimilation, producing . The first of the examples below is explicitly identified as an Eld (Common Eldarin) form in Etym.
mat > mat- to handle (Q mahta-)*2 *was > was stain (Q *wax > vax), 3

As the latter example indicates, eventually becomes k before s, producing ks = x. T H E L O S S O F M E D I A L ( OR H ) A N D Except initially, these sounds normally disappeared, though became before unvoiced consonants (see above) and survived before g and k (including labialized gw, kw the labialized sound w arising from + w had already become gw and thus preserved ). Much later, the language would re-acquire , developed from earlier g, only to lose it once again (even initially). The loss of original and this later has certain parallels, but whereas the later survived into historical times (it was recorded in Rmilian orthography, at least in initial position), the original or h was lost very early: Tolkien also referred to the loss of intervocalic h as prehistoric (VT39:11). Even more definite is the statement made in WJ:368: Medial h was very early lost without trace in CEtying this change to the Common Eldarin stage. A final following u might develop into w (or perhaps rather: the loss of triggered the development of a semi-vowel -w from the corresponding full vowel -u, as if to compensate for the lost ):
ku > kuw bow (Q c)

Sometimes, the loss of intervocalic and /h would not cause any further changes, but simply leave two vowels in hiatus; occasionally a new diphthong might arise after the loss, like ai in the second example below:
*te > *te > indicates (Q ta; cf. VT39:6) maiti > *maiti handed (Q mait)4 In Etym, entry MA, the CE word is spelt maht-, but it is clear that Tolkien is here using the letter h for rather than a simple breath-h; cf. Quenya mahta-, where it is clear from the descriptions in LotR:1087 that ht represents t. Again Tolkien's actual spelling of the primitive word is wahs.
*

COMMON ELDARIN

19

The primitive dual pe producing Quenya peu pair of lips would also be an example of a new diphthong arising after the loss of . Whether e became eu directly, or remained as two vowels in hiatus until the much later shortening of the final vowels, has only academic interest. The Quenya genitive ending -o is derived from the root HO after this loss of medial h (a primitive postposition h from being directly suffixed and turning into *-, later becoming -o).5 Two identical short vowels brought into contact by the loss of or /h would merge into one (long) vowel:
pee > p lip (VT39:11; still p in Quenya) maha > m hand (VT39:11; still m in Quenya)*

However, if the last vowel was long, and the loss of or brought it into contact with another vowel of similar quality (long or short), the first vowel might be dissimilated, like e > i :
te or *te > (*te >) *ti path (Q ti; see the entries TE, TE in Etym, the latter replacing the former)

Possibly the form *te survived for a while, the actual dissimilation to *ti taking place much closer to historical times: Similar dissimilation can be observed later as well, as when w produces Quenya vi manhood (via *w > *w). But in this case is not original; it is altered from g, since the root is not *WE, but WEG (LR:398). The change g > (and its subsequent loss) occurred much later. Where or disappeared before a consonant, the preceding vowel was lengthened in compensation:
dom > *dm night (Q lm)6 ter or ter > *tr straight, right (Q tra; again see TE, TE in Etym) war > *wr dirty (Q *wra > vra)7

It should be noted that while medial h/ disappeared, it surprisingly seems to have survived finally, being lost only much later in that position. This phenomenon is not explicitly referred to in Tolkien's published writings, but we shall probably need final later, to explain the Book Quenya accusative formed by lengthening the final vowel of a noun. This probably indicates the former presence of a final as a primitive accusative ending. But if we let final disappear already in Common Eldarin, it would be lost too early to exert any influence over the final vowels of later Quenya. So a word like *lass or *lassh, the probable primitive accusative of lass leaf, must remain unchanged for now. During the Common Eldarin stage, the short final vowels -a, -e and -o were lost. Examples:
*

These examples are interesting, indicating that the loss of medial and h occurred very early, since short final -a and -e were lost during the Common Eldarin stage; see below. Pee was evidently contracted to p before pee would have been reduced to **p (with the same lengthening of the vowel as in PQ kwene > CE kwn). While *p would probably also have produced Quenya p, this does not seem to have been the way Tolkien imagined it, since he referred to p as a contraction of pee (VT39:11).

20
The Evolution from Primitive Elvish to Quenya Loss of final short -a; where the ending -la is so reduced, the remaining l came to be syllabic:
swanda > *swand sponge (Q hwan with stem hwand-) jakta > *jakt neck (Q yat with stem yaht-) hekla > *hekl outcast (Q hecil) makla > *makl sword (Q macil) tankla > *tankl brooch (Q tancil) tekla > *tekl pen (Q tecil)

Loss of final short -e:

i-ndise > indis bride PQ kwene person > CE kwn (WJ:360) nere man > nr (WJ:393)

Loss of final short -o:

abaro refuser > abar (WJ:371 identifies the latter as a CE form) ndro > *ndr land (Q nr, WJ:413) Spanturo > *Spantur Cloud-lord (Q Fantur)8 tollo > *toll island (Q tol with stem toll-) Denwego > *Denweg (personal name, Q Lenw)9

As is evident from the examples kwene person > CE kwn (WJ:360) and nere man > nr (WJ:393) already quoted, a word that becomes monosyllabic by the loss of a final vowel has its stem-vowel lengthened in compensation. But where the word occurs compounded or with endings and thus is not a monosyllable, the vowel remains short. Therefore the CE plural of kwn is given as kwen (WJ:360), not **kwn. This variation is still reflected in Quenya qun pl. queni (WJ:361; likewise nr man pl. neri, MR:213). Where the loss of a final short vowel made a semi-vowel -w final, it turned into a full vowel u:
(angwa >) CE *angw > *angu snake (Q ango)*

A final semi-vowel following a long vowel could likewise develop into a separate vowel:
(PQ ?*rwa >) CE *rw > ru lion

This is our sole certain example of this phenomenon (Tolkien only provided the root RAW and the form ru). This -u was later lost, evidently also during the Common Eldarin stage. It is unclear whether we should generalize a rule a short final vowel immediately following a long vowel is lost:
ru > r lion (still r in Q)

After the loss of a final vowel by any of the processes described above, the plural forms were evidently reformed by analogy with the new simplex form. In Primitive Quendian, the
Compare the entry-head ANGWA in Etym with the stem-form angu- of the Quenya noun ango, seen in the compound angulc, entry LOK. Like -w became -u, a -j becoming final would likewise turn into -i; cf. the example talrunja sole of foot producing Quenya tallun, evidently via *talrunj and then *talruni; see below. In this case the loss of a final -a is later than CE, indicating that the rule for how final -w and -j behaved (turning into -u and -i) continued to operate in the earliest stages of Quenya. If the entry-head NOWO in Etym is taken as representing a PQ word *nowo, Quenya n conception, idea might arise in this way: PQ *nowo first becomes CE *nw (just like kwene becomes kwn), this *nw then turns into *nu, and the final -u is subsequently lost leaving only *n = Q n.
*

COMMON ELDARIN

21

plural of swanda sponge must have been *swanda or (later?) *swandai, but when swanda was reduced to *swand in CE, its plural was altered to *swand (still reflected in Q hwandi) as if this were a normal consonant stem. The original plural of PQ *rwa (or *rwe, *rwo) lion must likewise have been altered from *rwai (or *rwei, *rwoi) to *rw. Here, w was not final and therefore did not become u, later lost as described above. So in Quenya, the plural of r lion is still rvi, reflecting CE *rw and the original root RAW. Evidently during the latter part of the Common Eldarin stage, short final -i turned into - e:
dmi > *dme twilight (Q lm)10 pori > *pore flour, meal (Q por) *kari > *kare does (aorist form of the verb kar-) (Q car) *karini > *karine I do (aorist with pronominal ending; cf. Old Noldorin/Sindarin yurine I run, Q carin); see footnote below. *talruni > *talrune sole of foot (Q tallun)

The vowels did not change their quality when they were not final, and so in Quenya we still have the variation seen for instance in the aorist, as in car does besides carin I do (the first evolving from *kari with the change i > e, the second from *karini with a 1st person ending, the pronominal suffix *-ni precluding the change i > e because the vowel was not final).* As a parallel to the change from short -i to -e, short final -u becomes -o:
*kuru > kuro skillful device (in Q still curo with stem curu-; the examples below must also have stems in -u when endings are added to them) smalu > *smalo pollen (Q malo)11 tundu > *tundo hill, mound (Q tundo) ranku > *ranko arm (Q ranco) *angu > *ango snake (Q ango)

However, it should be noted that before the plural ending , words that earlier ended in gu and ku turned these combinations into the labialized sounds gw and kw (the sounded vowel u merging into a preceding velar consonant). The plural of *angu snake therefore turned into *angw (Q angwi). Likewise, the plural of ranku arm came to be rankw (Q ranqui). Compare WJ:390, where Tolkien indicates that the Quenya noun urco, pl. urqui descends as the plural form shows deom a form that originally ended in u rather than o: either urku or uruku.The plural urqui comes from *ur(u)kw. When the labialized velars gw and kw arose (before or after u turned into o in simplex forms) cannot be determined, since Tolkien had the stem-forms in u- surviving this shift anyway.
We date this change to the Common Eldarin stage because it is also seen in Old Noldorin/Sindarin: The form yurine I run (YUR) exemplifies an I-stem *yuri- with an ending -ne I, whereas the form trenare he recounts (NAR2) shows that the change from older *trenari has already occurred: Add the ending -ne I to trenare so that the final -e is not final, and we would evidently see *trenari-ne with the original quality preserved. The ending -ne I would itself exemplify the change from -i to -e, since it must come from earlier ni; compare the stem NI2, simply defined I. Another relevant ON/OS form is barane brown, which is surely meant to represent older *barani (cf. Quenya varn with stem varni-). However, the change -i > -e did not happen at the beginning of the Common Eldarin stage; this is evident from the fact that Tolkien refers to phini a hair [> Q fin] as a Common Eldarin form in PM:362 (cf. PM:340 for Q form). It may also be noted that Nandorin, Green-Elven, seems to have branched off from Common Eldarin before this change occurred; see the entry lygn in the wordlist appended to my article about Nandorin in the Ardalambion website.
*

22
The Evolution from Primitive Elvish to Quenya

EDITORIAL NOTES
1 But Helge brought one case to my attention. In VT44:14 we have the derivative (factitive) verbal form camta to (make) fit, suit, accommodate from the PQ base KAM fit, suit, agree. Helge noted that this form (descended from kam-t) is a rare, if not unique, case of primitive *-mt- not changing to -nt- as it certainly contradicts evidence from Etym. (Compare this with the base KAM as described in page 56, note 3.) As noted in p. 15 note 4 above the root MA in Etym was replaced by Tolkien to MAHA. According to VT39:11 note 6, this is a case of intervocalic h that was preserved before t. Even so, this would be a case of ht becoming t in Quenya. There are those (I am one of them) who contend that the spirant does not exist in PQ up to at least early CE, though it evidently existed during the late CE or pre-record Quenya stage as in the early Quenya form alda > later alda as well the Common Eldarin [root] OR (VT41:11). Therefore the root must have been *WAG. See note 2. The ulterior form should be *mahiti > *maiti > Q. mait. Cf. the primitive form magiti > *maiti > Q. mait shapely, seen in the name Maitimo (VT41:10). In Etym, the root is O. The later form HO is found in WJ:368. Based on a later source, the form dm is attested as the primitive form of Q. lm, S. d (Silm:365). See note 3. Tolkien replaced this with Fantur, and with a different etymology, but example is still valid within its purview. The form Denweg is still preserved in Nandorin. Cf. S. Denwaith, an alteration of Danwaith (= Q. Nandor) due to the influence of Denweg. See note 5. As implied in PM:366, Tolkien changed the root from which Q. malo is derived to *MAL instead of SMAL in Etym (cf. QL:58 s.v. MALA2). The form MALAT gold is the extended form thereof.

4 5 6 7 8 9 10 11

PRE-RECORD QUENYA
As far as the imaginary Arda timeline is concerned, the March is now past, and the Lindarin branch of Elvish (that leads to Sindarin and Amanya Telerin) has now definitely separated from the Vanya-oldorin branch (VT39:10) that we will continue to pursue here. The Vanyar and the Noldor have settled in Aman, but the linguistic development of this period still has to be reconstructed; Quenya still has not been reduced to writing. An actual word of what seems to be pre-record Quenya is found in WJ:402: Mbelekro is there said to be the oldest Q[uenya] form of the name Melkor. This word indicates that the primitive nasalized stops mb, nd, g were still in place initially, and that the syncope still had not taken place: Mbelek- would later become Melk-. Tolkien actually asterisked the form Mbelekro, so it is clear that Rmil still had not invented writing. Hence this can properly be called pre-record Quenya. It was in the period that now followed that the language underwent the greatest changes. Its altering ...[came]...in the softening and harmonizing of the sounds and patterns of the Quendian tongue to forms that seemed to the Noldor more beautiful (WJ:20). A long final vowel occurring at the end of a compound is normally shortened; Tolkien explicitly dated this change to the earlier stages of Quenya (VT39:6):
*Erukhn > *Erukhna Child of Eru (*Er, *Eru- God + khn child; cf. WJ:403 s.v. Atan) *tal-runj > talrunja sole of foot (*runj sole combined with tal- foot; Q tallun) *pel-taks > peltakse pivot (taks nail combined with pel- revolve on fixed point; Q peltas with stem peltax-)

Cf. also -wego as the compound form of weg man (LR:398 s.v. WEG). However, the plural ending - is not subject to this change (kala-kwend light-elves was not reduced to **kala-kwendi). It is also evident that many Quenya compounds were formed after this rule ceased to be effective for instance, Elentri Star-Queen would rather have been **Elentar if this descended from *Elentr at the oldest stage. One example mentioned above, *kemtan > *kentan potter, may likewise actually be a later formation (suggesting that the rule mt > nt continued to operate in the language): If this had been reduced to *kentano in early Quenya, the Quenya form might have been **centan instead of centano because of the loss of certain final vowels that we will now discuss. At some later stage, in words of three or more syllables, a short final vowel may occasionally be lost (but not wholly regularly so):
Awadelo > *Awadel Away-goer (Q Auzel > Aurel; this word refers to the Elves that departed from Beleriand and went to Valinor, WJ:360, 363; the form below is a variant) Awdelo > *Awdel Away-goer (Q Oazel > Oarel) Mbelekro > *Mbelekr Melkor (Q Melkor) (etsiri >) late CE *etsire > Q etsir mouth of a river (*karini >) late CE *karine > Q carin I do

24
The Evolution from Primitive Elvish to Quenya However, this loss seems to occur quite regularly in the case of the final vowels that had been shortened in compounds (another change dated by Tolkien to the earlier stages of Quenya, VT39:6):*
(*Eru-khn >) *Erukhna > *Erukhn Child of Eru (later Q Eruhin with stem Eruhn-) (*pel-taks >) peltakse > *peltaks pivot (later Q peltas with stem peltax-) (*khala-tirn >) khalatirno > *khalatirn fish-watcher (name of a bird, English kingfisher; in later Q halatir with stem halatirn-) (*Mand-ost >) Mandosto > *Mandost (later Q Mandos with stem Mandost-, MR:350 cf. VT39:6)

Short final -i at some point became -e just as in Common Eldarin, so it would seem that

this rule continued to operate in the language: *talruni became *talrune (later Q tallun). This parallels the change of -w > -u > -o in Common Eldarin (as when angwa snake becomes *angw > *angu > Q ango), and it is essentially the same rule continuing to operate in the language (but after this stage, it apparently ceases to function). Very early, there were some instances of metathesis, whereby certain sounds often switch place. One example is dl > ld, as in Edl > *Eld (Q Eldo, a variant of Elda Elf, though ultimately with another etymology see WJ:363). Tolkien wrote that transposition of dl in primary contact to the favoured sequence ld not infrequently [but not universally either?] occurred in the pre-record period of Quenya (WJ:363). Later, more systematic metathesis would occur. This may be the stage where Quenya loses many of the initial consonant clusters of the primitive language. The initial combinations sk-, sp-, st- evidently turned into aspirates kh-, ph-, th-, merging with the original aspirates (but medially these groups were apparently unchanged, cf. sk surviving in a Quenya word like rusco fox). This change is not directly attested in Tolkien's published papers, but we know he imagined such a change in the branch of Eldarin that lead to Sindarin. (In both Quenya and Sindarin the original aspirates and the initial combinations of s + k, p or t will later become spirants [> h], f, , respectively. However, Amanya Telerin retains the combinations in s- unchanged, so this change must have happened independently in Quenya and Sindarin: If it occurred already in Common Eldarin, we would have seen it in Telerin as well.)
skarw > *kharw wound (Q *arw > harw) skjapat- > *khjapat shore (Q hyapat) *skwarm > *khwarm crossbar (Q hwarma) spang > *phang beard (Q fanga) stank > *thank split (Q *anca > sanca)

The initial groups sl- and sr- become hl- and hr-, sc. unvoiced l and r:
Perhaps Mbelekro > *Mbelekr also belongs in this list, if the oldest Q form represents a compound *Mbelek- mighty + -ro, shortening of *r?rising or ?One who rises. The name Melkoras it later appearssupposedly means mighty-rising, also translated He that arises in might. In the scenario of Etym, original initial sl- seems to be simplified to l- at a relatively early stage; as one of the primitive derivatives of the stem SLIG we have lig (presumably representing even earlier *slig, cf. other primitive derivatives from the same root, slign or sling). In Etym, all Quenya words derived from stems in sl- show simple l-, which is of course logical enough if s early disappeared with no trace. (Etym provides no example of how initial sr- came out in Quenya.) But later, Tolkien decided that primitive initial sr- and sl- produced Quenya hr- and hl-, though this had come to be pronounced as normal r- and l- by the Third Age. So when Etym has a primitive word like (say) slw yielding Quenya lv instead of *hlv, we may see lv as representing the Third Age pronunciation. But in reality, Tolkien probably did not imagine any intermediate stage with hl- when he wrote Etym in the thirties; the introduction of hr- and hl- as Quenya
*

PRE-RECORD QUENYA
srw > *hrw flesh (Q hrv) *slk > *hlk dragon (Q hlc)

25

The initial groups sm- and sn- are simplified to m- and n- (conceivably there may have been an intermediate stage with unvoiced hm- and hn-, paralleling hl- and hr- above, but there is no direct evidence that Tolkien imagined this to have been the case):1
smalin > *malin yellow (Q malina) *snard > *nard knot (Q narda)

Perhaps it is at this stage certain primitive initial clusters in -j are altered. Primitive kj becomes tj (via tj?), primitive sj becomes hy (= ich-Laut), whereas primitive gj, dj and gj are all reduced to a simple semivowel j-:
kjelep > *tjelep silver (Q tyelp) sjalm > *hyalm shell, conch (Q hyalma) gjern > *jern old, worn (Q yerna) *djelt > *jelt to loathe (Q yelta) *gjalm > *jalm clamour (Q yalm)

The hw-sound (unvoiced w) enters the language as a simplification of initial sw-:

(swanda >) *swand > *hwand sponge (Q hwan with stem hwand-) swesta (read *swest) > *hwest to puff (Q hwesta) *swinj > *hwinj to swirl (Q hwinya, cf. the entry SWIN in Etym)

Another source of hw is the aspirate ph, where it occurred before the vowel u (cf. VT41:8):
*phuin > *hwuin gloom, deep shadow (Q huin)

BEGINNING OF THE RETRACTION PERIOD In primitive Elvish, it was not predictable (from the shape of a word) which syllable received the stress. In a little less than 30 cases, Tolkien in Etym indicated by means of an accent which syllable is stressed (plus some additional cases where he indicated which syllable in the root is stressed). For instance, it would seem that the primitive words bard lofty, berk wild and mork bear were stressed on the second-to-last syllable. But at some early point in the evolution of Quenya, the stress was moved to the first syllable in nearly all cases. (The only known exception is the augment, or prefixed reduplication, in verbal stems: according to WJ:366 this remained unstressed.) So now the three words above became *brad, brek, mrok but since practically all words are henceforth stressed on the first syllable, there is no point in marking the stress at all. The period from now and until the emergence of the later Latin-style stress-patterns we know and love from LotR-style Quenya is called the retraction period (WJ:366). It seems to cover the entire formative phase of the Quenya language, so the original stress-patterns of Primitive Elvish actually have minimal impact on the evolution of High-Elven. Exactly when Tolkien imagined the retraction period to have begun we cannot know, but it should be noted that the vowels that were originally accented in our three examples (bard, berk, mork), have all been lost in their Quenya descendants: Varda, verca, morco. They disappeared in the Quenya Syncope (see below). It is not probable that they
sounds was a later idea. In the pre-LotR period it would seem that he meant sl- to have been simplified to lalready in the primitive days. Cf. sm- and sn- below.

26
The Evolution from Primitive Elvish to Quenya would have been syncopated if they had still been accented, so the retraction period must at any rate have begun before the syncope set in. *Salkw grass must also have transferred its accent to the first syllable (*slakw) before it underwent syncope to produce the immediate ancestor of Quenya salqu. Shorter words that Tolkien indicated were accented on the final syllable, like the verb barj protect or the related adjective (or participle?) barn safe, protected were also stressed on the first syllable from now on. The Quenya descendant forms varya and varna are still so accented. THE QUENYA SYNCOPE Where a word of three or more syllables commences in two concomitant syllables that contain identical short vowels, the second of these vowels is normally syncopated (cf. Letters:426):
*doron > *dorn oak (Q norno) galad > *gald tree (Q alda) (karani >) late CE *karane > *karne red (Q carn) kjelep >) *tjelep > *tjelp silver (Q tyelp) *morok (no longer stressed on the second o!) > mork bear (Q morco) neres > *ners (Q Nessa, name of a Vali: WJ:416) (phoroti >) late CE *phorote > *phorte northern (Q fort) silim > *silm shining white (adj.) (Q silma) turum > *turm shield (Q turma) Ulum > *Ulm, name of the Vala of all waters (Q Ulmo) (salakw >) *salakw > *salkw grass (Q salqu)

However, the syncope did not take place if there was a consonant cluster before or following the second vowel. The word *aklar glory (Q alcar), where the second vowel follows the group kl, cannot be syncopated to **aklr. A word like *kjelepn of silver could not become *kjelpn, and therefore the second e still persists in Quenya telemna (altered from *tyelemna under the influence of Telerin, that had t for ty). A double consonant likewise hinders syncope in the previous syllable: tjuluss poplar-tree is not reduced to **tjulss (the second u is still present in Quenya tyuluss).* 2 A long vowel could not be syncopated, so a word like terw was not reduced to **terw (Quenya tereva fine, acute still preserves the second e; as we see it did become short
The fact that double consonants would prevent the syncope from taking place, confirms one sound-shift set out above: k + w had already merged into a single labial consonant kw. If salk-w (LR:385) had not become *slakw by this stage, there would still have been a consonant cluster following the second a, and the syncope producing the form *salkw (directly underlying Q salqu) could not have taken place. I cannot readily explain why atakw building (LR:390, later presumably *atakw) is not syncopated to *atkw ; the second a is still present in Quenya ataqu. But a form *atqu would probably be impossible in Quenya, and a combination like *tqu = *tkw would maybe be disliked also at earlier stages of Vanya-Noldorin; perhaps the syncope was sometimes avoided where certain undesirable consonant clusters would otherwise arise. Cf. also a word like tathar yielding Quenya tasar (willow) rather than syncopated (**tathar > **tare >) **tassThe syncope did not regularly take place in the branch of Elvish leading to Sindarin; hence primitive karani (late CE *karane, which would also be the Old Sindarin form) comes out as caran in Grey-elven. Where Sindarin words are syncopated, it may at least sometimes be the result of the more sporadic syncope taking place already in Primitive Quendian. Therefore we postulated above that (usuk-w >) *usukw reek became *uskw very early and did not have to wait for the Quenya syncope: Though the Quenya form would be usqu in any case, Noldorin/Sindarin osp demands a syncopated form *uskw existing already in Common Eldarin (in the Lindarin dialect of CE actually *usp).
*

PRE-RECORD QUENYA

27

later, but by then the syncopation had ceased to operate). Another example of the same is narka (read *nark), that was not reduced to **nark (Quenya naraca rushing, violent). Also contrast the development of two variant forms of the same word (WJ:360, 363): CE awa-delo away-goer becomes syncopated to *aw'del > *audel (Quenya Auzel > Aurel), whereas the variant aw-delo shows no syncope because of the long : This produces Quenya Oazel > Oarel, the group oa later arising from earlier awa (after it had been shortened from original aw, but too late for the syncope to take effect).* Where the syncope brought j or w into direct contact with a consonant, they became vowels i, u and could form a diphthong with the previous vowel:
gajak > (*gajk >) *gaik fell, terrible, dire (Quenya aica, PM:363) (awa-delo >) *awa-del > (*awdel >) *audel Away-goer (Quenya Auzel > Aurel) *tawarin > (*tawrin >) *taurin (Q taurina of wood)

At some early point (but after the final separation of the branches of Eldarin that lead to Quenya on the one hand and Amanya Telerin on the other, and evidently also after the syncope), the consonant d is changed in many positions. Initial d is consistently altered in the proto-Quenya branch. It normally becomes l:
(dmi >) late CE *dme > *lme twilight (still lm in later Quenya) datt > *latt hole, pit (Q latta)

If the word d'rk[] wolf still survived with its initial cluster dr- (produced by syncope of the original base DARK-) intact, it may have been simplified to *rk at this point. If the form *lrk ever existed, it hardly lasted long. The original initial cluster dr- has in any case been simplified in the later Quenya form rca, whenever the simplification may have occurred. Occasionally, initial d becomes n instead of l:
dr > nr man (still nr in Quenya) (*doron >) *dorn > norn oak (Q norno)

In the latter case, d evidently becomes n because of assimilation with the other n in the word. In WJ:413-414, Tolkien comments on the derivation of the Q adjective norna tough, stiff, which is implied to come from *dorn; he notes that this is probably one of the cases in which Q initial d became n-, not l-, by assimilation to an n occurring later in the word. (The Quenya words nenda slope and nend sloping, derived from a root DEN, seem to display the same assimilation: primitive *dend, *dend. However, Tolkien struck out the entry DEN [LR:354].) As for dr > nr man, Etym implies that this unexpected development is partly due to [influence from] N, NIS woman, partly to

The fact that the syncope only occurred in words of more than two syllables is responsible for the fact that a Quenya word like toron brother is reduced to torn- before an ending (e.g. pl. torni). The singular toron had only two syllables and was therefore unaffected by the syncope, whereas the three-syllable plural form *toron was reduced to *torn (> Quenya torni). But sometimes, inflected forms were later re-formed on analogy with the singular form, the syncopated vowel being reintroduced. The noun elen star is an example of this: The pl. form eleni, without syncope, is re-formed after the singular (WJ:262). But a syncopated form *eln had actually existed at an earlier stage, yielding an alternative Quenya plural eldi that might turn up in verse (for the change ln > ld, see below). Other Quenya examples of what seems to be analogical plurals are talami floors and filici birds (instead of syncopated **talmi, **filci; cf. syncopated nelci rather than **neleci for teeth).

28
The Evolution from Primitive Elvish to Quenya strengthened stem ndere bridegroom (LR:354 s.v. DER); therefore this word did not end up as **lr in Quenya (LR:377 s.v. N1).* Medial, intervocalic d may also become l by assimilation with another l in the same word, but this only happens sporadically:
g-lada- > *glala- verbal stem to laugh (Q lala-, PM:359)

Contrast lda (read *ld) as a primitive word for flat; this did not become *ll (which would have produced Quenya **lla, but we actually see lra, the consonant r regularly evolving from post-vocalic d see below). The primitive word *ld derived from a stem DAL seems to exemplify another phenomenon: Occasionally, the initial change d > l would cause an l occurring later in the word to be dissimilated to d. In effect, d and l switch places, as when the root DAL seems to have become *LAD in the derivative *ld (whence Q lra). Tolkien explicitly deals with another example of this phenomenon in WJ:363: by dissimilation, the root DEL go, proceed turned into led- in Quenya. (Etym only lists LED- go, fare, travel [LR:368]; when writing Etym, Tolkien had apparently not yet discovered that this is actually a dissimilated form of DEL-!) According to WJ:363, there is yet another way d can become l in Quenya: dj became ly medially. This is from the same paragraph that discusses DEL vs. LED, so we are obviously to assume that Quenya lelya go descends from older *ledj. However, this does not agree very well with one example from Etym: The root MAD yields a Quenya word marya pale, clearly meant to come from *madj (post-vocalic d regularly evolving into r via and z; see below). According to the rule Tolkien later set out, *madj should have yielded Quenya *malya instead, the cluster dj becoming ly before d had the chance to become r. But instead of altering the word, we may simply see marya as a lone sample of an irregular development. As we have seen, Tolkien could not quite make up his mind whether a specific primitive sound was to be reconstructed as h (presumably like English h), or as a back spirant . But if this sound was not h all along, the back spirant was changed to h at this stage at the very latest. We know this because Quenya was about to acquire from another source (changed from initial and post-vocalic g), and this sound would not behave like the older (? >) h. The latter survived into Quenya as h whereas the new would eventually be lost completely. Therefore, they cannot have been identical during the next stages of the linguistic evolution. THE SPIRANTS ARISE So far proto-Quenya did not have any spirant sounds (except possibly , which had by this stage at the latest finally become h, as noted above). But now a wide range of spirants would develop from two sources: the voiced stops and the unvoiced aspirates. Turning these sounds into spirants may be one of the changes Tolkien referred to as a softening of the older form of Elvish (WJ:20). Initial and post-vocalic g and b became and v, respectively:
(galad >) *gald > *ald tree (Q alda, but alda in the archaic Rmilian spellingsee below) But in a post-LotR source, Tolkien derived Quenya nr from a root NERE instead (WJ:393), eliminating the complications above: In the new scenario, the initial n of nr had been n all along.
*

PRE-RECORD QUENYA
*weg > we manhood (Q *wi > vi) (Denwego > *Lenweg >) *Lenweg > *Lenwe (personal name, Q Lenw, WJ:412) (lugni >) *lugne > *lune blue (Q ln, possibly *lune in Rmilian spelling) bes > *ves married pair (Q *vezu > veru) laib > laiv ointment (Q laiv)

29

It seems that g became following l as well:

ulgund > *ulund monster (Q ulundo)

It may be that not just g, but also b turned into a spirant following l (hence lb > lv), but the evidence is somewhat divergent. Etym has it that the cluster lb produces Quenya lw (as in golb > Q olwa branch, entry GLOB). But Tolkien seems to have abandoned this idea; in a post-LotR source the word for branch is quoted as olba instead (PM:340) as if the original cluster lb was unchanged in Quenya. However, LotR Appendix E3 informs us that for lv, not for lw, many speakers, especially Elves, used lb. So perhaps golb does properly yield Quenya *olva, but many speakers would alter it back to olba, and it is this alternative form that is quoted in PM:340. The cluster lv is not very common in Quenya, but in some of the cases where it does occur (e.g. in the word elva starlike), it almost certainly comes from earlier lb. We can probably feel free to postulate a regular development lb > lv, though many speakers, especially Elves would later undo this change apparently to distinguish this cluster more clearly from lw. Paralleling the changes b > v and g > , post-vocalic d became . The only direct evidence for this is Tolkien's statement (in WJ:363) that a primitive word edelo would have produced Quenya ello via elo; the word ello was not actually found, but the development d > must be valid all the same. Hence:
(*tada >) CE tad > *ta thither (Q *taz, in Exilic tar) (awadelo >) *Audel > *auel > auzel (Q Auzel > Aurel) (nidw > *nidw >) *nidwa > *niwa cushion (Q *nizwa > nirwa)

As for the first example, compare the primitive allative ending -da (WJ:366), hence *ta-da that-to = to that, thither. (Quenya mir to the inside; into must likewise descend from *mida via *mid > *mi > *miz.)* Possibly about the same point in time, the unvoiced aspirates also turn into spirant sounds. This seems to have occurred in all positions, though most of our examples involve initial aspirates. The change from aspirates to spirants seems to have happened more or less independently in all Eldarin tongues; for instance, Tolkien noted that kh became in all Eldarin tongues the spirant sc. the spirant , ach-Laut (VT41:9). Likewise, the aspirate ph became the spirant f, whereas th became the spirant :
(khthi >) *khthe > *e mist (Q * > hs) tathar > *taar willow-tree (Q *taar > tasar) Neth-r > *nera young (Q nessa)4 khjeles > *jeles glass (Q hyell) khithme (read *khithm) > *im fog (Q *iw > hisw) (thindi >) *thinde > inde grey (Q ind > sind, WJ:384) (phini >) *phine > fine a single hair (still fin in Q, PM:362/340)
*

No examples of initial d becoming can be quoted, since in this position d had already become l (or exceptionally n). Hence a primitive word like dom produces Quenya lm rather than (dom > dm > **me > **zme >) **rm.

30
The Evolution from Primitive Elvish to Quenya Since it seems that Quenya had already turned original initial sk-, sp-, st- into aspirates kh-, ph-, th-, they now became spirants just like the original aspirates (hence Tolkien was free to change his mind about whether find tress, hair comes from spind or phind: LR:387 vs. PM:362!)
(skarw >) *kharw > *arw wound (Q *arw > harw) (skjapat- >) *khjapat > *japat shore (Q hyapat) (skwarm >) *khwarm > *warm crossbar (Q hwarma) (spang >) *phang > *fang beard (Q fanga) (stank >) *thank > *ank split (Q *anca > sanca)

It should be noted that the that now arose in proto-Quenya was not quite like the sound of English th as in think. This English sound is interdental, sc. pronounced with the tip of the tongue between the teeth. The Quenya was strictly dental, made with the tongue-tip behind the back of the upper front teeth, and therefore closer to s (and in the Noldorin dialect, it would indeed merge with s later). Likewise, f was not originally quite like English f; the f of Amanya Elvish was originally a bilabial sound, pronounced using the lips only, without any friction between the teeth and the lower lip. It was quite close to hw (unvoiced w), and later in the Vanyarin dialect the sounds would indeed threaten to fall together. See VT41:7-8 concerning all this. The combination j, whether descended from khj or skj, becomes hy (= ich-Laut):
(khjeles >) *jeles > *hyeles glass (Q hyell)5 (skjapat- > *khjapat >) *japat > hyapat shore (still so in Q)

The labialized combination w becomes hw (= unvoiced w):

(*skwarm > khwarm >) *warm > hwarm crossbar (Q hwarma)

Because of this early merger of w with hw, distinct signs for chw [w] and hw were not required in Fanorian writing (LotR Appendix E)6. There really was a merger, for it should be noted that the sounds hw and hy were not quite new to the language: It seems that at this stage, proto-Quenya already had hw derived from sw7 and from the aspirate ph in front of u, and possibly also hy derived from earlier sj- (see above). THE SHORTENING OF FINAL VOWELS IN POLYSYLLABIC WORDS At some point perhaps not very far from the beginning of the historical period, the long final vowels so characteristic of Primitive Elvish are finally shortened. (The late Common Eldarin and early proto-Quenya rule that final short -i and -u turn into -e and -o is no longer operating, so the final -u and -i that now develop from earlier - and - are not so changed.)
korn > Q corna round Valinr (WJ:413) > Q Valinr Valinor tr > Q tri queen kund > Q cundu prince

Evidently by analogy, the shortened vowels were also introduced in positions where the vowels were not absolutely final, as in the inflected forms of nouns. For instance, the dative of lass leaf (earlier lass) is lassen, not **lassn.

PRE-RECORD QUENYA

31

The example Valinr from WJ:413 is of interest regarding the sequence of the changes: It demonstrates that the change from voiced plosives to spirants, like b > v, occurred before the shortening of the long final vowels.* However, Quenya still had some long final vowels at a later stage, in the accusative form of nouns. In the Plotz Letter, Tolkien indicated that in the conservative Book Quenya, a noun like cirya ship had the accusative form ciry, whereas lass leaf had the accusative form lass. How are we to account for these forms? It has been suggested that at an older stage, there was an accusative marker *-h. Whereas medial h was lost already in Common Eldarin (WJ:368), it may be that final -h survived much longer. If older lass leaf had an accusative form *lassh, the long vowel of the latter word would initially be shortened by analogy with the development of the simplex form: the new accusative would be *lasseh. However, final *-h was apparently lost before the beginning of the historical period, and the preceding vowel was lengthened in compensation. Hence presumably *lasseh > acc. lass, *ciryah > acc. ciry, long final vowels arising anew. (Presumably we would even see long final vowels in the accusatives of nouns that originally had short final vowels.) DEVOICING OF MEDIAL COMBINATIONS Perhaps at this point, a number of medial clusters of two voiced sounds are seen to become unvoiced. Examples include zd > st, zg > sk, and gd > kt:
(esd > ezd >) ezde > est rest (Q Est as name of a Vali, WJ:403) (mizd >) *mizde > Q mist fine rain (*masg > mazg >) *mazga > *maska soft (Q maxa) (khagda [read *khagd] >) *agda > *akta mound (Q *ahta > hahta)

Moreover, we can deduce that bd at one stage turned into pt:

(*libd >) libda > *lipta soap (Q lipsa)

We rely on the example libda to place the devoicing of these combinations in the period postdating the shortening of the long final vowels. The example khagda points in the same direction, but other evidence suggests that the initial aspirate kh- should already have turned into a spirant - when the final vowel had become short. The combination pt (of whatever origin) later turned into ps:
(libda >) *lipta > Q lipsa soap (*lepet >) *lepte > Q leps finger8

Tolkien may have imagined a different sequence when he wrote Etym, where we find some primitive words that already seem to have shortened the long final vowels, but still have not turned the voiced stops and the unvoiced aspirates into spirants. For instance, we have rba (> Q rva wild) and khagda (> Q *ahta > hahta mound) with b and kh still intact, though the final -a has already been shortened. It must have been long at the oldest stage, or it would not have survived into Quenya (cf. swanda > Q hwan sponge, not **hwanda). But then there are also words in Etym that would seem to confirm that the spirants entered the language before the final vowels were shortened. Cf. we (> Q vi manhood), that comes from a root WEG and must therefore have been *weg at an older stage; the form we cannot be Primitive Quendian or Common Eldarin, but is rather proto-Quenya, after the voiced stops had become spirants in many positions. Yet in we the long final vowel is still in place, confirming the order of changes suggested by Valinr in WJ:413. The material in Etym is thus not wholly consistent; we may simply assume that the macron denoting a long final vowel has dropped out in a number of cases. This is also seen in later sources; in WJ:361 a certain Primitive Quendian verb is quoted as hek-t, but in WJ:363 we have hek-ta instead plainly an error, whether the author or the transcriber is to be blamed.

32
The Evolution from Primitive Elvish to Quenya For the primitive form *lepet, cf. the root LEPET and the Noldorin cognate lhebed (Sindarin *lebed), LR:368. Conceivably the most primitive form should rather be *lepeti, but the immediate ancestor of Quenya leps would still be *lepte. Before the separation of Vanyar and oldor, both clans alter hw- to h- before the vowel u- (VT41:8):
(*phuin > *hwuin >) *hwuin > huin gloom, deep shadow

In Etym, this word has a Quenya side-form fuin (LR:382), but according to VT41:8, fuin is actually a Telerin form. Initially, the primitive nasalized explosives mb-, nd-, g- are normally reduced to m-, n-, and -. (As for certain alternative developments, as when ndn produces Quenya andn instead of **nn, see below.) If we can trust the examples ndeuna and ndeuro found in the entry NDEW in Etym, this simplification took place after the shortening of the final long vowels (or we would see *ndeun and *ndeur with the long vowels intact):
(*ndeun >) ndeuna > Q neuna second (*ndeur >) ndeuro > Q neuro successor (golod > *gold >) *goldo > Q oldo Noldo (*gwalm >) *gwalme > Q walm torment9 (mband >) *mbando > Q mando custody (MR:350) (*mbarat > *mbart >) *mbarta > Q marta fey, fated*

Properly, this reduction only occurs initially, but in some cases, it can be observed medially as well. The primitive word andambund long-snouted produces Quenya andamunda elephant (not **andambunda). The latter element of a compound might be reformed after the simplex (LR:350); in this way Tolkien in Etym explained how bal-ndre Valar-land could end up as Valinor instead of *Valindor: Occurring as a simplex, by itself, *ndr yielded Quenya nr land and what would be *Valindor was altered to Valinor by analogy (but see WJ:413, note 25, for a somewhat different explanation which Tolkien came up with decades later). So *andambunda might become andamunda by analogy with a related word like mundo snout (itself < mbundu). Where the second element of a compound normally does not occur as a separate word, nd/mb/g is normally preserved unchanged also medially. Prominent examples are the endings -ndil friend and -ndur servant, as in Eldandil Elf-friend (WJ:410). There are no separate simplex words **nil and **nur, though longer derivatives from the same stems are found in Quenya (e.g. nildo friend < *ndild). Indeed we may also include the element -ndor land, as in Valandor Vala-land (WJ:413, a source written after Tolkien had rejected or forgotten his original explanation of Valinor in Quendi and Eldar he invoked, not analogy with a simplex form, but rather confusion with a distinct root that never had initial nd- to explain why we do not see *Valindor). NEW VOWELS Around this stage, a number of vowels appeared out of nothing in Quenya. This usually occurred to turn syllabic consonants into normal syllables including a vowel. Henceforth, syllabic consonants were not tolerated in the language. This explains why in some cases the pre-nasalized initial stops mb, nd, g were not simplified to m, n, as set out above. As described above, during the Common Eldarin
*

For the middle vowel of *mbarat, later syncopated, cf. the base MBARAT and the Noldorin/Sindarin cognate barad.

PRE-RECORD QUENYA

33

stage the nasal element had occasionally been strengthened so that it had come to constitute a syllable by itself. Tolkien explicitly deals with one example in PM:360: The primitive noun golod could appear both with and without syllabic . With no syllabic it produced Quenya oldo as described above. But where a syllabic occurred (*golod), it came out as Ingoldo in Quenya, an i developing before the syllabic consonant. Ingoldo was a strengthened form, being in full and more dignified form is more or less equivalent to 'the oldo, one eminent in the kindred' (PM:360). As is evident from this example, an i developed before syllabic (except in the case of gw, see below) but before a syllabic n an a appears, and before a syllabic m there develops a u:
(ndn >) *ndne > andne sunset (still andn in Q) (*gj >) *gjo > *igjo grandchild, descendant (Q indyo) (*mbarta [final vowel uncertain] >) *mbart > *umbart fate (Q umbar with stem umbart-)

Syllabic before gw however prefers the vowel u (as we noted above, it seems that this had itself become labialized to w already in Common Eldarin, by assimilation to the following gw and its labialized quality colours the quality of the vowel that develops before it):
(*gwal >) *wgwale > uwgwale torture (spelt ungwal in Q)

Syllabic l triggered the development of the vowel i. As recorded in the entry LA in Etym, there was a negation prefix la- not-, un- that had lost its vowel (not a regular development), so that only a syllabic (or vocalic) l was left: This prefix became il- in Quenya, as in ilfirin immortal (for *lphirin-). It has been suggested that the Quenya verb ilca- to gleam comes from the same root that produces Sindarin lach flame, evidently *LAK-; if so ilca- could come from an original *lk, an unusual primitive form suppressing the stem-vowel.10 And i likewise developed before the final syllabic l's that had become final after the loss of final short -a in Common Eldarin. This change gives these words their final Quenya shape (the difference k/c is of course purely orthographic):
(hekla >) *hekl > Q hecil outcast (makla >) *makl > Q macil sword (tankla >) *tankl > Q tancil brooch (tekla >) *tekl > Q tecil pen

It may be that the Q word siril rivulet comes from *sirla > *sirl; if so, this would demonstrate that the vowel i developed before a final syllabic l before the assimilation rl > ll occurred (see below) the intruding vowel breaking up the cluster before it could be assimilated. In one case, an a rather than an i develops before l. In Etym, Tolkien derived the noun cala light from a primitive form k'l, perhaps = kl. But since the stem is given as KAL, it would seem very much easier to derive cala from *kal, instead of supposing a strange primitive form that suppresses the stem-vowel only to have an identical vowel arising later. In some cases an a also developed before r, because it had become syllabic in certain positions:
(*kwentr >) kwentro > Q quentaro narrator (the distinction kw/qu is of course orthographic only)

Here we have a syllabic r developing after a consonant cluster. In the case of Q tambaro knocker (used = woodpecker), which Tolkien indicates was ultimately derived from

34
The Evolution from Primitive Elvish to Quenya tamr, we must probably assume an intermediate form *tambro. There are no other examples of the strange development mr > mbr, but if a group mbr arose, the further development would confirm that r became syllabic following a consonant cluster. The word Istar wizard is perhaps to be derived from *istro > early Q *istr, given the Sindarin cognate Ithron (apparently representing an extended form *istro-n-; the point is that for st to have become th = in Sindarin, it must at an earlier stage have been in direct contact with the following r). CHANGES AMONG MEDIAL CONSONANT CLUSTERS Around this stage, various medial consonant clusters underwent changes. Some were complete assimilations resulting in double consonants:
LR > LL: (talrunja >) *talrune > Q tallun sole of foot (kalr >) *kalro > Q callo hero NL > LL: (*nenl >) nenle > Q nell brook LS > LL: (tles > *tels >) *telsa > Q tella hindmost (khjeles > *jeles > *hyeles > *hyels >) *hyelse > Q hyell glass RS > SS: (neres >) *Nersa > Q Nessa (name of a Vali) (WJ:416) SM > MM: (asmal >) *asmale > Q ammal yellow bird11 SR > RR: mi-srawanwe > *mirrawanwe incarnate (Q mirroanw, MR:350) R > : (neth-r >) *nera > *nea (Q nessa young)

Regarding medial sm, the examples are somewhat divergent. In asmal > Q *ammal we see a straightforward assimilation. Yet Tolkien (in LR:362 s.v. KAS) has primitive kas-sa/kas-ma yield Quenya cassa helmet; if cassa is to be referred to kas-ma, we must postulate a progressive assimilation sm > ss (and then ammal should have been **assal instead!) However, cassa is perhaps rather to be derived from kas-sa. In a later source, the Quenya word for helm(et) appears as carma instead (PM:260). If this is to be derived from kas-ma, the group sm is not assimilated to mm (or ss), but simply remains unchanged. (Later, s occurring between two voiced sounds is regularly voiced to z, in turn producing Exilic Quenya r hence the form carma. But then Q ammal from asmal should have become **armal instead!) It seems that we shall simply have to accept that medial sm could develop in various ways. In WJ:363, Tolkien also indicates that l would be assimilated to ll, though no good example can be quoted (Tolkien mentioned that a form elo would have produced Quenya ello, but this word he asterisked, for it was not found). The group r would likewise have been assimilated to rr, and here we may have a direct example, though a garbled one: From a primitive form njadr, literally *gnawer, Tolkien derived a Quenya word for rat that his son read as nyano (LR:379). However, this is certainly a misreading for nyarro, representing older *njaro. (In the normal Roman spelling of Quenya, nj is represented as ny, but this is only a matter of orthography, like kw = qu.) Since post-vocalic would later become r (via z) anyway, we cannot be certain that this assimilation really occurred. Otherwise, the word for rat would appear as *nyazro in the Vanyarin dialect. The example nenle (not older *nenl) given in Etym would seem to indicate that the assimilation nl > ll occurred after the shortening of the long final vowels, so perhaps all of these assimilations can be placed at this stage. There are also some cases of partial assimilation/dissimilation (not producing double consonants):
LN > LD: (skaln > *khaln >) *alna > *alda hidden (in later Q halda)

PRE-RECORD QUENYA

35

Cf. also WJ:362 about the plural form (elen > *eln >) *elni stars becoming eldi (though it was usually replaced by the analogical form eleni).*
PN and BN both > MN: (Utupn >) *Utupno > Q Utumno (Melkor's first stronghold) (stabn >) *abno > Q *amno carpenter (in later Q samno) (tubn >) *tubna > Q tumna low

The nasal m becomes w following n or :

(tinm >) *tinme > Q tinw sparkle, glint; star (khithme [read *khithm ] > *im >) *ime > Q *iw fog (in later Q hisw)

There are also two attested cases wherein bs becoming ps:12

(*labas? >) *labsa > Q lapsa to lick (frequentative) (abasen- >) *absen- > Q apsen- to forgive (VT43:18)

There is also a unique case of tf (< *tph) becoming ff:13

(*et-phir- >) *et-fir- > Q effir- expire (VT43:34-35; the actual attested form is effriemmo of our death)

Certain other changes involved metathesis, two sounds switching place:

KL > LK (LC): (*aklar >) *aklare > Q alcar glory (also short alcar) KR > RK (RC): (*lakr >) lakra > Q larca swift TR > RT: (netere >) *netre > Q nert nine (VT42:26; in this source Tolkien explicitly refers to changes TL, TR etc. > LT, RT)

Again, we have an example (lakra instead of older *lakr) that would seem to place these changes in the period post-dating the shortening of the final vowels. (The example akr narrow > Q arca [LR:348] would then skip over the intermediate form *akra.)14 Tolkien is seen to have changed his mind repeatedly about the behaviour of the medial cluster sk: Did it survive unchanged to become Quenya sc, or did it undergo metathesis to produce Quenya ks = x? In the entry EZGE in Etym (suggesting a primitive word *ezg) a Quenya word esc is listed; this sc represents the sk that arose after the regular devoicing of older zg. However, the entry EZGE was struck out. In other entries, Tolkien let sk (whether
It seems that occasionally, *ln may yield Quenya ll instead of (or as well as) ld: cf. ndulna > Q nulla or nulda secret (LR:355 s.v. DUL). Such attested past tense-forms as will flew and ull poured may then represent older *wilne, *ulne; cf. -n as a common past tense ending in Quenya. But it may also be that these past tense forms represent older nasal-infixed stems, so that the double ll seen in Quenya actually comes from nl rather than ln. In one exceptional case, bn becomes mb instead: The Quenya adjective lemba left behind is explicitly referred to lebn (LR:368 s.v. LEB). This must be due to an early, irregular metathesis lebn > *lenb > *lemb, perhaps helped by the alternative root LEM listed in the same entry in Etym (LEM could have produced a form *lemb already in the primitive language). In the case of samb room (LR:387 s.v. STAB), Tolkien listed both stabn and stamb as primitive forms, but these are evidently meant as variants co-existing already in the primitive language: While Quenya samb clearly comes from stamb , the Noldorin/Sindarin word thafn is to be derived from stabn ; the true N/S cognate of Q samb is actually tham listed in the same entry. Thus the meaning cannot be that stabn became or was replaced by stamb, so we have no second attestation of the development bn > Q mb after all. On the contrary, the entry STAB in Etym confirms the regular development bn > Q mn, since the Q word samna wooden post there listed is certainly meant to come from *stabn. Cf. also the adjective lumna burdensome, derived from a root DUB; the primitive form is clearly meant to be *dubn. The reconstruction Utupn is found in MR:69; Etym gives Utubnu instead [LR:394 s.v. TUB], but the Quenya form would be Utumno in either case.
*

36
The Evolution from Primitive Elvish to Quenya original or devoiced from zg) undergo metathesis to become Quenya x (actually spelt ks in Etym). Thus the word mazg (entry MASAG) yields Quenya maxa (maksa), and given the primitive form, this must be metathesized from an intermediate form *maska. The entry MISK indicates that original sk (not devoiced from zg) behaves in the same way; here we find a Q word mixa (miksa) wet, which given the entry-head must descend from *misk at the oldest stage. However, Tolkien seems to have changed his mind back again later: In a post-LotR source we find a word like rusca red-brown (VT41:10), which must be connected to the entry RUSK in Etym (in the later source, Tolkien actually struck out rusca, but retained related words like rusco fox).15 So now there is no metathesis sk > ks, just as suggested by the rejected entry EZGE in Etym. If we are to go with Tolkien's final (?) insights on this matter, the Quenya words mixa wet, maxa pliant, soft and max dough (< *misk, mazg, mazg) would have to be altered to *misca, *masca and *masc. Perhaps we may simply assume that there were some minor dialectal differences and divergent developments within Quenya, and accept all words as they stand. The group rg becomes ry:*
(targ >) *targa > tarya stiff

At some point after the shortening of the long final vowels (as is evident from the example ektele), the group kt becomes t, but we are now so close to classical Quenya that we will represent t as ht in accordance with normal Quenya orthography:
(et-kel >) ektele > Q ehtel spring, issue of water (hek-t >) hek-ta > Q hehta forsake (WJ:361, 363) (*khagd > *khakt >) *akta > *ahta mound (in later Q hahta) (jakta >) *jakt > *jaht neck (Q yat with stem yaht-) (nukt >) *nukta- > Q nuhta- to stunt (WJ:413)

In a few cases, this change caused confusion. The verb (makt- >) *makta wield a weapon had until now been distinct from (PQ *mat > CE *mat >) *mata to handle. But now *makta came to be pronounced just like *mata, and the verbs fell together in the form we would normally spell mahta in Quenya. Etym records how the verbs became blended (LR:371 s.v. MAK). The verb rihta to jerk may descend from either *rikt or rikht; Tolkien somewhat indecisively listed the root as RIK(H) (LR:383) either RIK or RIKH, or perhaps rather both. It may be that kh before t was reduced to k already in Common Eldarin; if so Q rihta would come from *rikt anyhow. But if *rikht survived in this form, it became Q rihta (=
In one case, Tolkien in Etym likewise lets lg become ly, Telerin felga cave having the Quenya cognate felya (PHLEG). The primitive form is clearly meant to be *phelg. However, this does not agree with another example, quoted above: ulgund monster does not yield Quenya **ulyundo, but ulundo (evidently via *ulund). We can probably afford to ignore the entire entry PHLEG in Etym, for it reflects an idea Tolkien later rejected: He is here explaining the origin of the first element of the name Felagund, which in Etym is Noldorin/Sindarin for Cave-Prince (cf. the entry KUND). However, Tolkien later decided that Felagund was not an Elvish name at all. In a post-LotR source it was re-explained as an adaptation of a Dwarvish (Khuzdul) term for Cave-Hewer: Felakgundu or Felaggundu and it is now actually the second element gundu rather than the initial element felag- that means cave or underground dwelling (PM:352). So from a certain point of view, Quenya felya cave and all its cognates must be rejected as Elvish words. It would seem, then, that Tolkien considerately eliminated a troublesome example for us. I would not let lg produce Quenya ly, but let lg behave as in ulgund > (*ulund >) ulundo instead.
*

PRE-RECORD QUENYA

37

rita) already in the general change kh > and did not have to wait for the change kt > ht (= t). Eventually (of whatever origin) before t would turn into ich-Laut [] following the vowels e and i; cf. the later Quenya pronunciation of words like nehta spear-head or tihta to blink. We cannot know at what stage this change happened (since it is not reflected in normal Quenya orthography anyway it may well be late). In any case, remained unchanged following the vowels a, o, and u; cf. the later Quenya pronunciation of words like mahta to handle/wield, ohta war, luhta to enchant. Various combinations of voiced stops + m also underwent changes. Etym quotes nakma (rather than older *nakm) as the origin of the Quenya word nangwa jaw; if we can trust this piece of information, this change also occurred after the shortening of the final long vowels. (In the case of a word like sungwa drinking-vessel, said to be derived from sukm, we may assume that Tolkien simply did not mention the intermediate stage *sukma with the final vowel shortened.)
KM and GM both > NGW (technically GW): (*nakm >) nakma > Q nangwa jaw (sagm >) *sagma > Q sangwa poison TM > NW: (*mbaratm >) *maratme > Q maranw destiny (jatm > *jatm >) *jatme > Q yanw bridge (the difference j/y is of course orthographic only)

As is evident from the example *tinme > Q tinw discussed above, *nm also turns into nw. By one theory, tm first became nm and then nw (*maratme > *maranme > Q. maranw); if so, the change *nm > nw actually happened after this stage. Like tm, the group dm would probably also become nw, but we have no examples. Probably already in the pre-record period, initial w is lost before long (WJ:367). Hence the prefix w-, meaning together referring to the union of two things, was reduced to - in Quenya (when unstressed shortened to o-), and we have for instance *w-mentie > Q omenti meeting (of two), *w-nna > Q adj. onna twin-born (WJ:367). Post-vocalic s followed by a voiced sound (any vowel or a voiced consonant) becomes z. (This z will eventually become r in Exilic Quenya.)16
(thaus > *aus >) *ausa > *auza foul (Q *auza > saura) (bes > *ves >) *vesu > *vezu married pair (Q *vezu > veru) (besn > *vesn >) *vesno > *vezno husband (Q *vezno > verno) (*pisj >) *pisya > *pizya syrup (Q *pizya > pirya)

The voicing s > r does not take place if s is followed by an unvoiced consonant (including another s, in the case of double ss):
(usuk-w > *uskw >) uskwe = Q usqu reek (not **uzqu > **urqu) (bess > *vess >) vesse = Q vess wife (not **vezs > **vers)

The evidence is somewhat divergent regarding the behaviour of a (single) post-vocalic s occurring finally. In Etym, Tolkien let s become z (> later r) in this position as well, as when he derived the Quenya noun solor surf from older solos (implying the existence of an intermediate form *soloz). The first word listed in the entry LOS is likewise olor (dream); the imagined development would be *olos > *oloz > olor. However, this very word reappears in UT:396, but now in the form olos though its plural is given as olozi/olori. Hence Tolkien has altered the rules for what precise environment the voicing s > z occurs in. In Etym, the rule is that a post-vocalic s becomes voiced unless it is followed

38
The Evolution from Primitive Elvish to Quenya by an unvoiced sound. The later rule seems to be that s is voiced if it occurs between two voiced sounds (either two vowels, or a vowel and a voiced consonant). Thus olos is still olozi in the plural, where the plural ending makes s intervocalic so that the voicing can occur.* The spirant merges with the newly developed sound z (and will also end up as r in Exilic Quenya):
(Awadelo > *Audel >) *auel > auzel (Q Auzel > Aurel) (nidw > *nidw >) *niwa > *nizwa cushion (Q *nizwa > nirwa)

When writing Etym, Tolkien imagined that non-initial t became s before the vowel i (at what point in the imagined timeline this change was supposed to have occurred is difficult to say; probably relatively late, since there is no trace of such a shift in Noldorin/Sindarin). Therefore, the adjective (maiti >) mait handy, skilled had the plural form maisi, ultimately < *mait. (From this example it is clear that this shift must have occurred after final short -i became -e in late Common Eldarin, or t would have become s in the singular form as well.) In the case of a cluster tt, only the second t changes to s before i, as when quelet corpse (stem quelett- < kwelett-) has the plural queletsi. However, it may be that Tolkien dropped this sound-shift after completing Etym. In a much later (post-LotR) source, Tolkien wrote hlonti tengwi, not ?hlonsi tengwi, for phonetic signs (WJ:395). So perhaps the plural form of mait, quelett- could simply be *maiti, *queletti as well. OTHER VOWEL CHANGES (INCLUDING DIPHTHONGS) Sporadic losses of final vowels still occur; some adjectives in -ina are shortened:
(*melin >) *melina > melin dear (*skwarin >) *hwarina > hwarin crooked (*phirin >) *firina > firin dead (*kwalin >) *qualina > qualin dead (by unnatural cause/s)

The Quenya descendant of *latin Tolkien listed as latin(a), apparently meaning that latina and shortened latin coexisted. The unattested longer forms of the adjectives above may also still be valid. Most adjectives in -ina seem to persist in their full forms (calina light, culina flame-coloured, enwina old, malina yellow, etc. etc.) Evidently before the beginning of the historical period, the diphthongs ae, ao and ei turn into long vowels , , and , respectively. (Ae and ao did not change directly; Tolkien implied that between ae and there was an intermediate stage, which he repreIf we were to impose full consistency on the material, we should have to alter some of the words in Etym. Solor surf would become *solos (to go with olos rather than olor), and seler sister (derived from older *elez, cf. the base THELES) would have to become *seles instead. Teler Telerin Elf, in Etym derived from TELES, would have to become *Teles in the singular, though the plural could still be Teleri for older *Telezi (cf. olos having the pl. olozi > olori in UT:396). However, Tolkien in a post-LotR source implied that Teler comes from *Telero instead (WJ:371), so this word can stay as it is (the r now being original, not altered from z < s). Perhaps Tolkien changed the etymology precisely because he did not want the singular to be *Teles (as it would have to be according to his Revised Voicing Rules). We could follow Tolkien's example and keep seler sister as it is by assuming that it actually comes from primitive *thelere (with a feminine ending *-re corresponding to masc. -ro added to *THELE, the extended form of the root THEL listed in LR:390 of which THELES listed in the same place is merely another possible extension). Concerning a curious example from the entry KAS, cr head with stem cas-, see the Appendix. We should probably read *cs with stem *car-.
*

PRE-RECORD QUENYA

39

sented as , and before ao turned into this diphthong first became a special A-like version of the vowel [represented by ]. These intermediate stages may have been short-lived.) Where one of the older diphthongs ao or ei (no examples for ae) had occurred at the end of a word, the language now regained some long final vowels in polysyllabic words. They survived throughout the historical period in Valinor and were directly recorded in the classical Book Quenya orthography (these long final vowels became short only after the Noldor went to Middle-earth). Ae becoming long :
(*mael >) *maela > mla loving*, 17

Ao becoming long :
(*kaol >) *kaolo > clo burden, 18

Also notice that in Book Quenya, nouns in -a are seen to have genitive forms in -, e.g. ciry as the genitive of cirya ship. The genitive form ciry arose at this point in the linguistic evolution and represents *kirjao, in turn derived from older kirj-h or *kirja-h. The vowel of the suffixed element -h from, of came into direct contact with the preceding vowel a after the very early loss of medial h treated above (WJ:368). Thus a final diphthong -ao arose.
Ei becoming long : (neiti >) *neite > Q nt moist (mein >) *meina > Q mna adj. eager to go (VT39:11)

Final -ei also becomes -, as in the plural of words ending in -e (the diphthong arising when the plural ending -i is added). The plural of lass leaf, that had long been *lassei, now becomes lass the form recorded in Book Quenya orthography. In one exceptional case, Tolkien let ei become ai instead of : In Etym (entry WEY), the name of the Vali Vair is referred to primitive weir, and there is a reference to a change wei > wai as if the preceding w has something to do with this unusual development. However, perhaps this example should be ignored. In a later source Tolkien derived Vair's name from an A-infixed form of the stem WIR (evidently meant to be *Wair), in which case the diphthong ai would simply be unchanged since the primitive period. (VT39:10) Virtually all final consonant clusters are simplified (the ending -nt for dual dative is the sole known final cluster in later Quenya); the clusters are however preserved before endings, since they are not then final:
(jakta > *jakt >) *jaht > Q yat neck (but with stem yaht-) (peltakse >) *peltaks > Q peltas pivot (with stem peltax- = peltaks-)

Final -st becomes -s:

(*Mand-ost > Mandosto >) *Mandost > Q Mandos (with stem Mandost-, MR:350 cf. VT39:6)

Final clusters in -d lose this consonant, e.g. -nd becoming -n:

(swanda > *swand >) *hwand > Q hwan sponge (with stem hwand-) Tolkien mentioned no primitive form, but this Quenya adjective is implied to be derived from an A-infixed form of the stem MEL-; see VT39:10. However, in Etym he derived Q. vr bond, troth, compact from wd, which in turn must come from *waed, with an A-infixed form of the root WED. This indicates that *ae first became and then a regular long in Quenya. Again, Tolkien in VT39:10 implies that an A-infixed variant of the stem KOL underlies this word.
*

40
The Evolution from Primitive Elvish to Quenya
(*t aurondo >) aurond > Q *auron (Exilic Quenya Sauron, quite possibly with stem Saurond-)
h

For the latter example, see Letters:380. Another example involves a simpification -ld > -l: In the Etym scenario, the name Taniquetil has the stem Taniquetild- because the latter element of this name is reduced from tild horn. The final -e was lost early, but the name must have appeared as *Taniquetild at one stage. Final -ng (actually -g) loses the -g and becomes a normal n. The name of the Golden Tree Laurelin was sometimes taken to mean Hanging Gold and assigned the stem Laureling-, the latter element being related to linga hang, dangle. To the extent this interpretation was correct, the name may indeed have appeared as *Laureling (i.e., *Laurelig) at one earlier stage. Final -m and -k become -n and -t, respectively. Thus the older base TALAM is the source of the Quenya word talan floor, with stem talam-. A word for bird derived from the stem PHILIK had by now probably reached the form *filik; in classical Quenya we find filit, with stem filic-. Cf. also nelet tooth from an original base NEL-EK (in this case the stem-form has undergone syncope: nelc-). Final double consonants are simplified, but remain before endings:
(tollo >) *toll > Q tol island (stem toll-) kwelett > Q quelet corpse (stem *quelett-) (*nisse >) *niss > nis woman (stem niss-)*, 19

EDITORIAL NOTES
1 However, it is my opinion that Tolkien had later rejected roots that begin with SM- and SN-. An example of this is the base SMAL, which he changed to *MAL (see p. 22 note 10 above); he definitely rejected the base from which the Noldorin word naith was derived in Etym (LR:387 s.v. SNAS; see UT:282 note 16 for its later etymology). As it can be suggested in a later source (SD:419), the ulterior form of Q. tyuluss may actually be *kjuluss instead. Helge commented, in an internal perspective, I guess different Loremasters disagreed about the ulterior forms of certain words, and various views are reflected in the Professors writings. That is, LotR:1095. According to a later source (WJ:404, 416) the name of the Vali Nessa was said to be Valarin in origin. However, there are those who contend that the name descended from primitive neres she that has man-like valor and strength; this sense may not have survived in later times. The Q adjective nessa young exists independently of the name, and as the original meaning seems to be almost forgotten (except among the Loremasters), Nessa came to be called the Ever-young. See page 14 note 2. That is, LotR:1097. Helge, in a later email communication to me, sees sw as an actual combination, though David [Salo] may see it as a unitary, labialized consonant (sw). But primitive pt was retained in Q raccalepta claw-fingered in SD:72. With thanks to Helge for bringing this attention to me through email. I would like to add tapta impeded in VT39:17. Helge afterwards surmised that Tolkien may have eventually dropped the entire ps > pt rule. If so, then the

3 4

5 6 7 8

The simplex nis woman also occurs in the form ns; it is difficult to account for the long vowel, unless this form is influenced by the archaic word n of similar meaning.

PRE-RECORD QUENYA

41

9 10

11 12 13

14 15

16

17 18 19

Quenya word for finger would have been *lept instead of leps. However, in VT44:16 in a note dated c. 1968, the Quenya word for finger is said to be leper, pl. leperi, and lepta a derivative verb meaning pick (up, out) with the fingers (< *lep-t). Spelt ngwalme in LotR:1096 I have to disagree with this proposed etymology of ilca. I dont think this was derived from *LAK, but rather from a class of roots which end with *-IL, all with the sense of shine or gleam like NGIL (see MR:388), RIL, SIL and THIL Helge later explained to me that this was actually David Salos speculation, and he agreed with me in principle. But see note 1above. This is a late addition suggested by Helge. A similar case can be found in Sindarin: ephel outer fence < *et-pele. But in this case I consider this change as unexpected, as it was assumed that *t-ph- would descend to *tf- in Quenya. Besides, I believe that this form was but ephemeral, as Tolkien changed the line s ar lmess effriemmo now and at the our of our death to s ar lmesse ya firuvamme (lit. now and at the hour that we will die) in a later version of Aia Mria. This should not be confused with Q arca pray, which is speculated to be derived from RAK (< *arak < *a-rk-). See VT43:33. The ultimate base, of course, is (U)RUS (VT41:10). The entry RUSK in Etym should be seen as a primitive word, not as a root. Helge explains, [I]t may be due to Christopher Tolkiens eager regularization of his fathers manuscript. It may be then that it should be written as RUS + k, but Tolkien struck out rusca in VT41:10. But according to VT44:20 note 4: The change of intervocalic s > r is not universal: compare, for example, sanwe. An unpublished bote from the 1960s states that this changes did not occur when s was followed by a stressed vowel (as presumably in sanwe, stressed snwe). Another example, not included but would explain its existence, is the attested from aselye with us (VT43:29), hence aslye (the accent here is an indication of stress, not length). Similarly, as suggested in VT39:10, the ulterior form of Q mla is *ml. The hypothetical intermediate ulterior form would be *kl > kl (thus S caul). The plural form nsi is also attested (VT43:30), although this may be an ephemeral form by Tolkien. However, the presence of the long vowel may be inconsequential, as there are cases wherein Tolkien was consistent about this in his language creations.

CHANGES PRODUCING VALINOREAN QUENYA

Probably before the beginning of the historical period, a double ii (evidently two vowels in hiatus) was simplified to i. The old perfect form a-wwii has passed away (WJ:366) would have become *awwiie with a short final vowel by now; perhaps around this stage, it was further simplified to *awwie. (In later Quenya this yielded avni, which is not a regular phonetic development; the n replacing the last w intruded by association with the past tense ending -n. See WJ:366.) At some relatively late stage, ai/ay (but not y with a long ) becomes e before an unaccented vowel (or at least before o and a; examples for other vowels seem to be missing). Probably ay first became ai; then it was reduced to e () in this position.
(daio [read *daj] > *laj >) *layo/*laio > lo shade (gajr > *ayr >) *ayr/*air > *r sea (cf. WJ:400; in Exilic Quenya ar) (phaj >) faya/*faia > fa spirit

The latter word is the first element in the name of Fanor, Quenya Fanro; the fact that it is quoted in the form Fayanro in PM:343 would seem to indicate that aya > a is a late change (at least post-dating the change ph > f). We do not asterisk the form faya, since it is attested in this name. It may be that Tolkien meant this change to have taken place during Fanor's lifetime, since his name could also be quoted as Fayanro. Paralleling this change, the group aw (not w) before an unaccented vowel is reduced to o (in WJ:367, Tolkien hints that aw first became au and then o):
(aw >) *awa > ua (WJ:367, the accent probably indicating stress rather than length) > oa away (WJ:366, VT39:6) (*law >) lawa > *laua > loa year (VT42:10) (sraw >) *hrawa > *hraua > hroa body (MR:349-350) (mi-srawanwe >) *mirrawanwe > *mirrauanwe > Q mirroanw incarnate (MR:350)

Strangely, Tolkien has the primitive word tawar wood (as material) yielding Quenya tavar instead of **toar (LR:391 s.v. TAWAR; as for w becoming v in this position, see below). In a later source, a word for wood is however given as toa, regularly derived from a root TAWA (VT39:6). The TAWAR of Etym could be simply an extended form of it. At some later point, when the rule turning ay/ai and + a vowel into e had ceased to function, the combination j (or in Quenya orthography y) became ai. The older forms with y seem to have survived into the historical period, so the Quenya designation of the lesser spirits of the people of the Valar can be cited as Myar (PM:363,364). Yet the word normally appears in its later form Maiar (which did not turn into **Mar since the older rule had ceased to operate, as noted above). Another example of the same is waia envelope, ultimately derived from wj (Etym, entry WAY).1 Perhaps some words escaped this change; in Etym, Tolkien cited the Quenya adverb far off as hya without mentioning any later form haia, but it does occur in SD:312.*

The word must ultimately be derived from *khj, closely reflecting the root KHAYA. In Valinorean Quenya it would actually appear as *ya; see below concerning the shift > h in Exilic Quenya.

CHANGES PRODUCING VALINOREAN QUENYA

43

The retraction period ends. The classical stress-patterns are established: Two-syllable words are nearly always stressed on the first syllable; longer words are henceforth accented on the second-to-last syllable if that is long; otherwise the stress falls on the third syllable from the end. At some point after the establishing of the new stress-patterns, a long vowel immediately preceding or following the accented vowel is usually shortened (but not universally so as far as a preceding vowel is concerned; for instance, in WJ:368 Tolkien notes that the Quenya prefix h- from usually retained its long vowel even when it had become unstressed). Shortening of vowels immediately preceding the accented vowel:
(*w-nn > *-nna >) *nna > Q onna twin-born (*mand >) *mandi > Q omandi pl. vowels (but singular *mand- > Q man with the long vowel intact since it is stressed)*

Many three-syllable words do not behave quite like we might expect. Even where there is a long vowel in the second-to-last syllable that ought to attract the stress, the accent remains on the first syllable even after the end of the retraction period, and the long vowel in the following syllable is shortened:
(Awdelo > *Awel >) *Ozel > Q Oazel away-goer (later > Oarel) (Art >) arta > arata exalted (PM:363, VT41:9; Q pl. Aratar used of the chief Valar) (*nark >) narka > Q naraca harsh (terw >) *terwa > *terewa fine, acute (> Q tereva) (tern >) *terne > Q teren slender

Perhaps we should also include ontaro parent here, if the accompanying form ontro listed in Etym (entry ONO) is to be taken as an older form rather than a mere alternative. (It is indicated that ontro was at some point accented on the second-to-last syllable, but during the retraction period, the stress would have fallen on ont-.) This development seems to be relatively regular when we are dealing with an originally long vowel in the penultimate syllable of three-syllable words: The vowel is shortened and the stress settles on first syllable. Perhaps *w-nn behaves differently (producing Q onna instead of **nona) because the initial syllable is a prefix so that this did not count as a quite regular three-syllable word? However, there are certainly further inconsistencies in the material on this point: The primitive form gl (> *gl > *ingle) produces Quenya ingol philosophy (PM:360), with the stress moving to the first syllable and the long vowel in the penultimate syllable becoming short, just as in naraca, teren and the other examples just listed. Yet primitive ndn (>*ndn > *andne) yields Quenya andn sunset with the long vowel in the penultimate syllable intact; this must also be the syllable that receives the stress. These two examples ought to be completely parallel (incidentally even in the fact that the initial vowel developed from a syllabic nasal), so why do they develop differently? If ingol, why not **andun? And if andn, why not **ingl (and **narca, **terva, **tern etc.)? It would seem that the outcome of the phonological evolution is not entirely predicable in this regard. Exceptionally the long vowel in the penultimate syllable attracted the stress and remained long, but in most three-syllable words the stress moved to the first syllable and the long vowel of the penultimate syllable became short.
*

The plural omandi appears as amandi in LR:379 s.v. OM, transparently a misreading or typo.

44
The Evolution from Primitive Elvish to Quenya On the other hand, another development has the stress moving towards the end of the word, a vowel being lengthened to accomplish this. A word ending in three short syllables would normally be stressed on the third syllable from the end. However, it seems that this stress-pattern is only tolerated in words that only consist of three syllables in all, so that the accented syllable is also the first syllable of the word. If the stress would normally fall on the third syllable from the end in a word of more than three syllables, and this syllable is short, the vowel in the second-to-last syllable is lengthened so as to attract the stress. Thus if we add the possessive ending -va to Eldali Elf-people, the result is not **Eldalieva (accented on the short syllable li), but Eldaliva (accented on v). Cf. also Cuivinen, which is cuivi awakening + nn, -nen water.* The adjective vanima beautiful, if supplied with the partitive plural ending -li to express beautiful ones, must of course become *vanimli. This is only attested in the genitive form vanimlion (occurring in Treebeard's greeting to Celeborn and Galadriel). The first edition of LotR had vanimalion, but this rejected form would conflict with the apparent rule that a noun of more than three syllables cannot end in three short syllables. Some compounds (like vanimor monsters, which is - un-, not + vanimor beautiful ones; or the name Altariel/altariel = alta + riel, PM:347) seem to be exempted from this rule. On the other hand, we observe the excepted lengthening in Anrion sun-son = Anar + the patronymic ending -ion.2 Normally, classical Quenya does not seem to permit a long vowel immediately before a consonant cluster. Mnwen or Mnw is quoted as the oldest Quenya form of the name Manw (WJ:399); notice how was eventually shortened before the cluster nw. Likewise, Friel's Song indicates that the noun m hand appears as mannar in the (plural) allative perhaps for older *mnnar. However, long vowels seem to be allowed before clusters ending in the semi-vowel y: Namri has mryat (rather than *maryat) for her hands. Some forms published in VT44 do to some extent complicate the picture. Vnde occurs in one text as the word for maiden; elsewhere (as in Etym, entry WEN-) the form used is vende. The editors suggest that vnde contains an anomalous and probably unintentional long vowel arising by confusion with the distinct word Vne, from which vnde was emended (VT44:5, 7). Moreover, Tolkien in one text used Hrsto (but also Hristo) as a Quenya adaptation of Christ (VT44:12). Since this is not meant to be a native Quenya word anyway, it may be only partially adapted to regular Quenya phonology. The normal rule does seem to be that there cannot be a long vowel in front of a consonant cluster. ANTICIPATORY VOWELS DEVELOPING FOLLOWING INITIAL + CONSONANT It seems that in some cases, a vowel developed following an initial consonant cluster in (representing earlier g-). This vowel seems to echo or rather anticipate the first vowel occurring after it, even where that vowel was the first element of a diphthong. Tolkien indicated that original grauk- underlies Quenya rauco and arauco terrible creature (WJ:415). Evidently grauk- (the complete primitive word must have been *grauk or *grauku) first became *rauk- with the general change g > . This would later disappear
Curiously, the combination ui seems to count as two short syllables for this purpose, as is this were a combination of two short vowels in hiatus rather than the diphthong Tolkien stated ui to be. Thus the possessive form of huin shadow is attested as huinva (LR:382) rather than **huineva, and a compound of tuil spring + r, -r day must appear as tuilr (mentioned in LotR:1082) rather than **tuiler.
*

CHANGES PRODUCING VALINOREAN QUENYA

45

(see below). *rauk- might either persist in this form, in which case it would simply be reduced to rauc- in Quenya, or it might develop an extended parallel form *arauk-, the vowel developing after anticipating the next vowel in the word (in this case the first element of the diphthong au). This *arauk- would later become Quenya arauc-. Etym seems to include one example of this phenomenon, though Tolkien later struck out the relevant entry: The stem GLINDI produced Quenya ilin pale blue (evidently shortened from *ilind). The idea must be that GLI- first becomes *li- and then *ili- with an anticipatory vowel, eventually reduced to ili- with the later loss of . Another possible example was apparently also rejected by Tolkien: In MR:284, the place-name Avathar (evidently = Avaar) is said to be ancient Quenya for the Shadows. This must somehow be related to Sindarin gwath shadow (See the entry gwath in the Silmarillion Appendix.).3 In Etym, this word is derived from a root WATH, and the prefixing of g- in the Noldorin/Sindarin branch is secondary. However, Tolkien may at one point have considered deriving gwath from a stem *GWATH instead, the initial G being original and therefore also able to influence the branch leading to Quenya. In this alternative scenario, the word shade, shadow might be *gwath in the primitive language, later becoming *waa > *awaa in the proto-Quenya branch, in turn producing archaic Quenya *avaa (see below concerning the change of post-vocalic w > v). Avaar could obviously be the plural of this word, so that the word could refer to the Shadows as Tolkien stated. This derivation is not entirely unproblematic, though: Instead of *GWATH and *gwath we would expect *GWATH and *gwath at the primitive stage; the initial combination would not strictly be a cluster g + w, but simply gw = labialized g, a unitary consonant. Later we would then have w (rather than w). Hence would not really be part of an initial consonant cluster, the environment that triggers the development of an anticipatory vowel. Could the labialized quality of w really be separated from it to materialize as a separate consonant w > v, an intruding vowel splitting an originally unitary consonant in two (w > aw-)? In Etym, the stem GWEN produces Quenya wend > vend maiden, not **evend for older *wende. (Cf. also g-lada- > *glala- > *lala- simply becoming lala- in Quenya, PM:359 but the hyphen inserted by Tolkien in the form g-lada- may indicate that the prefixing of g- was peculiar to the Lindarin branch, so that the branch leading to Quenya never had anything but simple initial l-.) It seems that Tolkien eventually gave up trying to find an Elvish etymology for the name Avathar, explicitly denying that it was Elvish and instead suggesting that it was a borrowing from Valarin (WJ:404). But since our source for this is the same essay that provides the example grauk- > Q arauc-, the idea of anticipatory vowels as such was evidently not rejected. The semi-vowel j is lost when it occurs after the corresponding full vowel i and before another vowel. W is likewise lost between the corrresponding full vowel u and a following vowel:
oij >) *oija > Q oia ever* (kukw > *kw >) *kwa > Q cua dove

But in later Exilic Quenya, the form *oiya = *oija might arise once again, because a glide consonant y occasionally developed between i and another vowelsee below.

46
The Evolution from Primitive Elvish to Quenya Tolkien mentions kua as the immediate ancestor of cua, so it would seem that the w of kwa became u before it simply merged into the preceding u. Also notice how the long becomes short when brought into contact with a following vowel. Such shortening occurs in other parallel examples from Etym (see below concerning tuo, ri). In a monosyllabic word, a final w merging into a preceding u does produce a long , still surviving as Quenya :
(ku >) kuw > k bow (= Q c)*

Older w, where it still survived after the various changes discussed above, in very many cases became v. It seems that the change of medial w to v occurred before the same change happened initially (the initial change w > v is rather a feature of Exilic Quenya see below). But post-vocalically, w normally became v very early, and it seems that this change would have taken effect already during the Classical Valinorean period. See further discussion below (under Exilic Quenya). Also notice that while Tolkien still thought of Avaar as an Elvish name meaning the Shadows (before he turned it into a Valarin word), he stated that it was ancient Quenya (MR:284). The v of Avaar certainly reprensents earlier w (this goes whether Avaar is derived from a root like WATH shade, cf. LR:397, or from *GWATH > *wa- > *awa- as we theorized above). Hence, the change of post-vocalic w to v had already occurred in ancient Quenya, perhaps meaning pre-record Quenya. Examples of post-vocalic w becoming v:
(srw > *hrw >) *hrw > hrv flesh (kuiw >) *cuiw > cuiv awakening (*rw >) *rwi > rvi pl. lions (cf. root RAW) (terw > *terwa >) terewa > tereva fine, acute

However, it would seem that w remained unchanged following the vowel o; cf. nwi as the plural form of n conception (the pl. form more closely reflecting the original root NOWO). For some reason, w also persists unchanged following the diphthong ai. In Etym, Tolkien has slaiw yielding Quenya laiwa sick (not **laiva). Contrast the related word slw, producing Quenya lv sickness; here there is no diphthong ai to hinder the development w > v. Other examples:
aiw bird (< aiw, given as an entry-head in Etym) maiw gull (< *maiw , with an A-infixed variant of the root MIW) vaiwa wind (< *waiw, cf. root WAIWA)

Depending on its position, the original back spirant had either turned into h (or ) or been lost, as described above. Now Quenya had re-acquired (developed from original g), but this sound was evidently not destined to be a lasting part of the Quenya sound-system: This new , whether initial, medial or final, also disappeared:
But the ultimate ulterior form may have been *kugu < a stem *KUG. For the most primitive form, cf. the prefixed variant et-kuiw listed in Etym, entry KUY. Since the Quenya possessive ending -va evidently descends from *-w, one may wonder why we see Eldaiva rather than *Eldaiwa as the pl. possessive of Elda (WJ:368). The explanation seems to be that -va had originally only occurred in the singular (Eldava, with no preceding diphthong ai); the plural form Eldaiva was a later, analogical formation based on the singular Eldava. See WJ:407.
*

CHANGES PRODUCING VALINOREAN QUENYA

(*weg > we >) *weo > *wo man (in later Q vo) (Denwego > *Denweg >) *Lenwe > Lenw (personal name, WJ:412) (gass > *ass >) *assa > assa opening (grauk[] > *rauk >) *arauko > arauco terrible creature (WJ:455) (gard > *ard >) *arda > Arda (a name of the world, WJ:402) (*gond > *ond >) *ondo > ondo rock (galad > *gald > *ald >) alda > alda tree

47

In some sources Tolkien only speaks of how primitive initial g- was lost in Q (Letters:410), but g did not disappear directly; it first turned into . We know this because survived long enough to be explicitly recorded in Rmilian writing: In Rmilian orthography, the word alda was still spelt aldathat is, it had an initial letter which was known by tradition among the loremasters to have represented the open back spirant (VT39:7). Hence we need not asterisk the form alda. (If we knew more about Rmilian writing, we could probably have removed the asterisk from words like *assa, *arda, *ondo as well.) Between vowels, and following l, the spirant is lost with no trace:
(*khugan >) *uan > *uan hound (Exilic Quenya huan) (ulgund >) *ulundo > ulundo monster

Where a postvocalic was followed by a consonant, the preceding vowel was lengthened to make up for its disappearance (maintaining prosodic length see SD:420 for Adnaic parallels):
(lugni > *lugne >) *lune > ln blue (*rign > *rigna >) *rina > rna crowned (sagr > *sagra >) *sara > sra bitter (magr > *magra >) *mara > mra good

It would be interesting to know whether Rmilian orthography represented these words in a spelling similar to the (bold) asterisked forms above.* It should be noted that a w that became intervocalic only by the loss of , missed the normal change of intervocalic w > v; this change had ceased to operate when disappeared:
(jagw > *jagwe >) *yaw > yw ravine (NOT > **yv)

According to Etym, a long vowel is shortened if, by the loss of , it is brought into direct contact with another vowel:
(tgu > *tgo >) *to > tuo muscle (not *to)
*

The noun huan hound (descended from khugan via *uan) is said to appear as hn- when endings are added. This would seem to come from *un-, exemplifying the same phenomenon of vocalic lengthening. It may be that soon after g became , an unstressed non-final syllable a lost its vowel: *uan- > *un- (> ultimately hn-). This did not happen where no ending followed so that the combination a was part of the final syllable; hence *uan by itself produces Quenya huan with no loss of a.

48
The Evolution from Primitive Elvish to Quenya
(rg > *rge >) *r > ri crown (not *r)

But according to a post-LotR example (PM:347), such a vowel remains long; this is actually a variant of ri above:
(rg > *rga >) *ra > ra wreath (not *ria)

Whether we should emend Quenya tuo, ri above to *to, *r is unclear. Perhaps we may rather assume that they actually descend from *tugu, *rig instead, the stem-vowel being short from the beginning. From one example it may seem that two identical vowels brought into direct contact by the loss of merge into one (long) vowel: (ugu- >) *uu- > -, the Quenya negative prefix un- (in Etym derived from a stem UGU). However, it may be that the second u was lost already in the syncope, and that *u-, usually occurring before a consonant, turned into by same process that turns *yaw into *yw (see above). In other examples of two vowels of the same quality coming into contact by the loss of , the first of them is dissimilated: o > o'o > uo, and ee > e'e > ie. (Compare e' becoming i in Common Eldarin, after the loss of intervocalic or .)
(Kh-gor heart-vigour >) *or > *o'or > *uor (masc. name, Exilic Quenya Huor, Sindarin Huor)* (*weg > we >) *wee > *we'e > *wie manhood (Q *wi > vi)

By a change that first occurred in Telerin, the original bilabial f became a labio-dental f, similar to English f (pronounced with friction between the lower lip and the upper teeth). This change made it easier to distinguish f from hw, so the Noldor adopted labio-dental f from Telerin (VT41:7). The Vanyar did not, and the Noldor accused them of confusing f and hw. In the dialect of the Noldor, unlike that of the Vanyar, nd is simplified to n before y (WJ:360, 361):
(kwendj >) Quendya > Quenya

SECONDARY SYNCOPE The last recorded syncope operating in Quenya involved the omission of the second of two identical short vowels where a word occurs in long compound names (VT41:9). We call this secondary syncope because it is unrelated to the much older, primary syncope that produced, say, al'da tree from original galad. The primitive word art had produced Quenya arata exalted, noble with the second a surviving the primary syncope because it had originally been long (PM:363, VT41:9). Yet a word like arata could be
We have again a long vowel being shortened before another vowel after the loss of ; the long evidently becomes short o and is then dissimilated to u before another o. According to the post-LotR example rg > *ra > ra quoted above, we might expect the vowel to remain long: *or > *or, later *Hor. A double form like endya, enya middle (Etym, entry NED) may be taken as the Vanyarin/original and the later Noldorin form, respectively. However, this would mean re-interpreting Etym material in the light of Tolkien's later ideas, for when he wrote Etym, there was no such thing as a Noldorin dialect of Quenya (the Noldor were still conceived as speaking something very much like the language Tolkien later called Sindarin). In the case of indyo grandson (entry GY), no later form *inyo is mentioned.
*

CHANGES PRODUCING VALINOREAN QUENYA

49

shortened to arta (PM:354) when it occurred as part of a longer name (e.g. in Artaher noble lord, PM:346, or masculinized -arto in Ambarto, PM:354).4 We must assume that the silma- part of the word Silmaril has also undergone secondary syncope, since this element represents the name silima that Fanor gave to the substance from which [the Silmarils] were made (the Silmarillion Appendix, entry sil-).*, 5

EDITORIAL NOTES
1 2 But the isolated form va from the name Vantur (see UT:470 for page references) seems to be likewise descended from waj. Helge later commented: Since Anar Sun descends from anr- (LR:348), it may be that in Anrion an old long vowel is simply preserved long because it happened to be stressed. The simplex sun would still have been Anr in earlier forms of Quenya (SD:306): according to what Tolkien wrote in the Plotz Letter, the shortening of long vowels in the final syllable of polysyllabic words is a feature of Exilic Quenya. See below. Lukas Novak had a similar observation way before in Elfling message 7706 (the first part of which is given in page 10 note 3): I think this fact could better be explained differently. There is not only altariel, but also Eldali itself, vanimor, and perhaps some words [that] I am not aware at the moment, whereas Alatriel though perhaps irrelevant (since [it is] Telerin) shows lengthening. I think [that] the explaining rule could be this: [m]ore than [three]-syllable proparoxytona lengthen the stressed syllable, if that is short and the preceding syllable is also shortthus avoiding an overlong sequence of short syllables. If I am right, this supports the first of my objections [see page 10 note 3]: according to this rule, [the example] avaquete to forbid [would have its stressed syllable, i.e. the second, lengthened], but it does not, since it is already long, which shows that the initial sound quet- is not an altogether unitary consonant, but a semicluster. Or you should have to assume that [the] one and the same qu is once pronounced as a cluster and once as a [u]nitary soundbut there seems to be no phonetical reason to distinguish one unitary sound into the original sound in [the] compound. It seems much more plausible to me that (some) semiclusters were allowed initial sounds in Q, and that somewhat different rules were applied on them than on ordinary consonant clusters. That is, Silm:359. Add to this the attested examples Artafind and Artanga, the proper Quenya equivalents of T. Findarto and Angarto, respectively (PM:346). That is, Silm:364.

3 4 5

In the earlier scenario recorded in Etym, the initial element of Silmaril is suggested to be the adjective silma silver, shining white, which is referred to a primitive adjectival form silim that would have lost its second i already in the primary syncope (LR:385 s.v. SIL). Indeed a primitive form of the word Silmaril is provided, namely silimarille (LR:383 s.v. RIL, the final - of silim is regularly being shortened when the word occurs as the first part of a compound). But the actual existence of such a primitive form is problematic for historical reasons: Surely the primary syncope was long since over when Fanor invented the Silmarils? It must be placed in the unrecorded past long before he was even born. Perhaps the form silimarille is to be seen as an entirely hypothetical back-formation even within the fictional context? But if we accept the new scenario set out in the Silmarillion Appendix that the first element of Silmaril represents a contemporary noun silima coined by Fanor himself we can appeal to secondary syncope regarding its reduction to silmain a compound. Thus the linguistic anachronisms are avoided.

CHANGES PRODUCING EXILIC QUENYA

The noontime of Valinor is drawing to an end. Soon, Fanor will lead the Noldor into exile, and Quenya will become known in Middle-earth as well as in Aman. In Middle-earth it will become a language of lore, largely shielded from further changes by the fact that it is learnt from writing by each new generation. But before the language to some extent becomes fixed like this, a few more changes will occur some of them greatly controversial, intertwined with Valinorean politics! They will define the shape of Exilic Quenya, by definition the only form of Quenya known in Middle-earth. These changes also defined the Noldorin dialect of Quenya, as opposed to the Vanyarin dialect (though it is possible that Vanyarin did not share with Noldorin all the changes so far listed). An important source of information about the changes affecting Exilic Quenya is the list of Tengwar names in LotR Appendix E.1 The original name of all the relevant Tengwar includes the sound this letter originally represented. But its precise pronunciation was later altered, as when l (in LotR Appendix E spelt thle)2 came to be pronounced sl instead. In some cases, a Tengwa could be renamed altogether because of the phonological changes. Wrote Tolkien regarding the Tengwar names, Where there are variants this is due to the names being given before certain changes that affected Quenya as spoken by the Exiles. This should not be taken to mean that all of these changes necessarily occurred during the Exile, after the Noldor had actually left Valinor. We know that Tolkien imagined the shift reflected in the change l > sl to have occurred in Valinor during Fanor's lifetime. The shift > s is indeed the most celebrated of all Quenya sound-changes, and Tolkien was able to develop an entire story around it: It was underway in Valinor during Fanor's lifetime, but he and all the loremasters deplored it, since it caused a great deal of homophony, and confused the derivatives of originally quite distinct stems. (VT41:8 actually there are very few words in our corpus that would merge in form; *amna > samna wooden post vs. original samna diphthong may be one example, but these are hardly difficult to distinguish in context!)3 However, Fanor's advice was presented with such arrogance and haughtiness that he rather accelerated the merger of and s: Into the strife and confusion of loyalties in that time this seemingly trivial matter, the change of to s, was caught up to its embitterment, and to lasting detriment to the Quenya tongue. Had peace been maintained there can be no doubt that the advice of Fanor, with which all the other loremasters privately or openly agreed, would have prevailed. But an opinion in which he was certainly right was rejected because of the follies and evil deeds into which he was later led. He made it a personal matter; he and his sons adhered to , and they demanded that all those who were sincere in their support should do the same. Therefore those who resented his arrogance, and still more those whose support later turned to hatred, rejected his shibboleth (PM:335-336). Hence:
l > sl spirit erind > Serind (the name of Fanor's mother; the change annoyed both mother and son, PM:333) ind > sind grey elma > *selma will (WJ:319) *iw > *isw fog (later hisw) riil > *risil circle

CHANGES PRODUCING EXILIC QUENYA

nau > naus imagination

51

According to Etym, final -s might also represent earlier - sometimes: The name Tulkas is apparently derived from Tulkatho in LR:395 s.v. TULUK; if so its older Quenya form might be *Tulka.* In Etym, there even seems to be an example of double ss representing earlier double *; the word nessa young is ultimately derived from neth-r, presumably via *ner > *nea. In Etym, this is presented as the origin of the name of the Vali Nessa. But in a later source Tolkien suggested that this name comes from neres instead, or explained it as a borrowing from Valarin (WJ:416, 404). However, this neres does not mean young as did neth-r, so perhaps the two must now be concidered wholly separate words, the later forms Nessa and *nea coinciding only in late Quenya, because of the /s merger. The change > s eventually became universal in Noldorin Quenya (in Middle-earth, even the sons of Fanor seem to have adopted the new pronunciation after his death), but the distinction was never dropped in writing (PM:332). According to LotR Appendix E,4 the letter now called sl continued to be used for s derived from earlier though we have an example of Tengwar calligraphy where Tolkien himself seems to have forgotten this, using the normal s-Tengwa (silm) even where he ought to have used sl. It should be noted that the change from to s did not occur in the Vanyarin dialect of Quenya, but only in the Noldorin dialect (thus sind rather than ind is said to be a [oldorin] dialect form in WJ:384). When Indis, herself a Vanya, married Finw, she made a conscious decision to change her pronunciation: I have joined the people of the oldor, and I will speak as they do (PM:336). Fanor was not amused. Another change that seems to have occurred already in Valinor, but very late, was the shift z > r in nearly all positions. (One exception: according to WJ:413, medial z escaped this change if there already was an r in the same or an adjacent syllable. If z thus did not become r, it apparently became s instead, for z was never used in Exilic Quenya.) The Tengwa that Fanor had originally named z (sunshine) thus came to be called r, but since there already existed another letter for r, r ceased to have any clear function: It would seem that the distinction between original r and r changed from z was not upheld in writing. (In Middle-earth, the now vacant letter r was eventually assigned a new value, double ss, and the letter itself was renamed ess to reflect this.) As in the case of the shift > s, the Vanyar were conservative; Vanyarin never shifted z to r, so the Vanyar would perceive this change as a dialectal feature of Noldorin Quenya. A place-name like Ezellohar indicates that z remained a living sound in Amanya Elvish, though the Noldor
However, Tolkien would later explain the name of this Vala as a borrowing from Valarin: WJ:399,404. As discussed earlier, was itself derived from the aspirate th, which in turn was sometimes original and sometimes changed from the initial cluster st-. However, it seems that when Tolkien wrote Etym, he imagined the change from to s to have occurred in the remote past not as late as during Fanor's lifetime. This is evident from a note in the entry STAR, where Tolkien observed that the Noldorin word tharn (meaning sapless) was not in Q that is, had no Quenya cognate since it would coalesce with sarn. Tharn comes from *starn (the Old Noldorin form stharna is given), whereas sarn comes from a distinct root SAR and produces the Quenya adjective sarna of stone. According to the later scenario, *starn would become *tharn and then *arna, remaining distinct from (sarn >) sarna until the merger /s occurred. Since Tolkien did not say that the words did coalesce so that one of them was suppressed, but implied that one had never been in Quenya because it would have coalesced with the other, we can infer that the merger /s was originally imagined to have taken place already in the pre-record period. One remark in the entry SIL, to the effect that this stem and THIL in Q cannot be distinguished normally, points to the same conclusion. Later, Tolkien moved the merger /s forward in the imagined time-line and had it occurring in Fanor's day.
*

52
The Evolution from Primitive Elvish to Quenya probably came to say *Erellohar instead. In WJ:363, the forms Aurel and Oarel (evidently Noldorin Quenya) are contrasted with Auzel and Oazel in the Vanyarin dialect; these Vanyarin forms also reflect the older, common Vanya-Noldorin pronunciation. This also seems to confirm that the Noldorin change of z > r occurred already in Valinor, or there would be little point in using it to contrast the Vanyarin and Noldorin dialects of Quenya. The z's that now become r may derive from either d or s in the primitive language:
(*olos >) olozi > olori pl. dreams (sg. olos, UT:396) (nidw > *nidw > *niw >) *nizwa > nirwa cushion

With this change, the already very limited number of consonants that Quenya can have finally is further reduced by one; after this change, only -t, -l, -n, -s and -r (original or derived from -z) remain. The letter harma Fanor originally meant to represent ach-Laut, the spirant , and undoubtedly the original pronunciation of its name was *arma. However, LotR Appendix E records how [x] became breath h initially (though remaining [x] medially).5 Hence:
(kher > *er >) *eru > heru lord (cf. Letters:282 and VT41:9) (khugan > *uan >) *uan > huan hound (skarw > *kharw > *arw >) *arw > harw wound

The early Quenya forms in - are implied to have existed, but no such form is directly quoted in the published material, so we must asterisk them. In Etym, only the Exilic Quenya forms in h- are recorded (that is, what would be Exilic Quenya forms in Tolkien's later scenario). It may be that Tolkien originally held this change to have occurred further back in history, in the pre-record period. But in the linguistic scenario as it appears in the LotR appendices, this weakening of initial to h can perhaps be ascribed to Sindarin influence on Quenya as spoken by the Exiles in Middle-earth: Sindarin had also turned initial into h. However, Tolkien noted that this change occurred in all Eldarin tongues (VT41:9), which would include the Telerin of Aman; hence this change was not unheard of in Valinor either. Perhaps it was already beginning in Valinor shortly before the Exile, but occurring too late to be reflected in the original Fanorian orthography. The Tengwa harma was therefore renamed aha (wrath), since in this word the middle consonant was still pronounced . Perhaps this is why *aa, as it must originally have been pronounced, is nonetheless spelt aha in Appendix E: In the end, medial also came to be pronounced h in most cases. The exceptions are probably the combinations normally spelt aht, oht, uht, where orthographic h ever remained a full spirant . In the combinations eht, iht it turned into ich-Laut, but in these groups it was quite possibly still represented by the letter aha in Tengwar orthography though we have no examples. When initial became h, it seems that it merged with the initial breath-h that already existed in Valinorean Quenya. This breath-h was in most cases derived from primitive (or, if Tolkien was in that mood, it had been h already in the primitive language and never changed e.g. Primitive Quendian hekla outcast > Quenya hecil in WJ:361, 365). Exceptionally h was also derived (via hw-) from the aspirate ph before u, as in huin gloom. Initial was on the other hand derived from the aspirate kh- (original or changed from even older sk-). The change - > h- in Exilic Quenya lead to some cases of homophony. The word harw treasury, representing primitive *arw (root AR, LR:360) would have had breath h already in Valinorean Quenya. But the word harw wound was

CHANGES PRODUCING EXILIC QUENYA

53

not originally a homophone; this comes from primitive skarw, which (after first becoming *kharw) yielded Old Quenya *arw. Only with the change > h did the words coincide in form.* The name of the Tengwa wilya was pronounced vilya in Exilic Quenya. It would seem, then, that initial w became v (merging with v derived from earlier b). Above we have discussed how w became v medially, and one may well ask why we cannot let initial and post-vocalic w become v at the same time but there is good reason to believe that post-vocalic w normally became v much earlier than the initial shift occurred. Consider this: The loss of must be placed very early in the historical period, since it was only by the tradition among the loremasters anyone knew what sound had once been represented by the initial letter of such a Rmilian spelling as alda (VT39:7). Yet the medial shift w > v must predate the loss of : In the case of (jagw >) *yaw > yw ravine a that was later lost hindered the development of post-vocalic w to v. A later shift w > v, after the loss of , would have turned *yaw > yw into yv (producing a homophone with the word for fruit, itself < *jb). So the evidence might seem to be contradictory: One line of reasoning leads to the conclusion that the shift w > v cannot be later than the earliest parts of recorded history, whereas Tolkien elsewhere seems to imply that w and v remained fully distinct at the time Fanor devised the Tengwar the sounds merging only in Exilic Quenya. The only way of reconciling all the evidence is to distinguish between a shift of post-vocalic w > v, which took place very early before the loss of (and which *yaw > yw thus escaped), and an initial shift w > v, which took place only after Fanor invented the Tengwar and is reflected in the change wilya > vilya. Since Erendil is said to have named his ship Wingalt or Wingelt (PM:370, 371) with initial w still intact, this change may have occurred after the end of the First Age. It would seem, then, that changes like the following happened among the Exiles in the Second Age or later:
wra > vra private, personal, own (PM:340, where it is made clear that this change occurred after the stage Tolkien called Old Quenya) waia > vaia envelope wn > vn goose waiwa > vaiwa wind wend > vend maiden There is one problem; in Etym Tolkien derived the word harma from a root AR, and then it should have had breath h already in Valinorean Quenya. Yet Tolkien picked this word as the name of the Tengwa originally denoting , which would imply that it was earlier pronounced *arma instead. We could assume that earlier breath h merged with in Valinorean Quenya, later reverting to h in Exilic Quenya (when original also became h). However, LotR:1097, footnote, indicates that even the original Fanorian system had distinct signs for breath h (originally represented by the sign halla, a raised stem with no bow) and the spirant (represented by harma). This would suggest that h and coexisted in Valinorean Quenya. Perhaps Tolkien simply changed his mind about the derivation of harma treasure, deriving it from a stem *KHAR instead of AR, so that it would originally have had initial . If the word halla (tall) contained breath h already in Valinorean Quenya, Tolkien may have altered its derivation as well: It is not actually listed in Etym, but it would probably have to be referred to the root KHAL2 as a cognate of Noldorin/Sindarin hall exalted, high (older form khalla given). But halla with original breath h would require a stem *AL (or, *HAL). Conceivably, halla tall simply describes the form of the letter (a raised stem) rather than exemplifying the sound it denoted: If the root is to remain KHAL, this word must have been pronounced *alla in Valinorean Quenya.
*

54
The Evolution from Primitive Elvish to Quenya Where Tolkien quotes Quenya forms in w- only, as he sometimes does, we must assume that he intends them to be Old Quenya, First Age Quenya. Etym is quite inconsistent in this regard: Quenya words derived from bases in W- are sometimes quoted with initial w-, sometimes with initial v-, and sometimes both. All contradictions are resolved when we recognize the forms in w- as Old Quenya, whereas the forms in v- are later. (In one case, vilin I fly derived from the root WIL, Christopher Tolkien interprets his father's notes to mean that he changed it to wilin: LR:399. However, there can be no question of vilin being rejected in favour of wilin; for some reason, Tolkien decided to quote this verb as an Old Quenya form rather than the modern Quenya variant.) It is not clear whether the distinction between w and v was upheld in Tengwar writing (the letter vilya still being used for initial v even after it had been changed from w), or whether all initial v's came to be spelt with the letter vala regardless of derivation. (Vala originally represented the v's that were present already in Valinorean Quenya; when initial they were always derived from older b-.) In his RGEO Tengwar transcription of Namri, Tolkien did use the letter vala to spell the initial v of the word vanwa lost, though this is understood to represent primitive *wanw and must have been *wanwa in Old Quenya. (But then he also used silm to spell the s of hsi mist, though this s represents earlier and should have been spelt with the letter sl according to the rules he set out elsewhere!) Another change implied by the Tengwar names involves initial becoming normal n (this includes labialized w becoming nw, normally spelt nw in Roman Quenya orthography). In LotR Appendix E,6 Tolkien records how the Tengwa originally called ngoldo (alternative spelling of oldo) had its name changed to noldo, and ngwalme (that is, walm) similarly became nwalm. In the same Appendix, Tolkien noted that the sound of (or ng) also occurred initially in Quenya, but has been transcribed n (as in Noldo), according to the pronunciation of the Third Age. Etym hints at a similar change. In the case of armo wolf from the root GAR(A)M, the later form narmo is also listed. In other cases, Tolkien is quite inconsistent about whether he lists the older form in - or the later form in n-; he may even be inconsistent within the same entry in Etym, as when the root GAW yields auro werewolf next to naul wolf-howl: When naul had reached this form (from older *aul), auro had likewise become *nauro! In the case of some words that are obviously related to walm > nwalm torment in LotR Appendix E, namely nwalya to pain and nwalca cruel from the root GWAL, only these later forms are recorded in Etym: Given the form of the root, the words must earlier have appeared as *walya and *walca. In the Plotz letter, Tolkien mentioned certain changes that occurred before Quenya ceased to be a birth tongue among the Noldor, hence early in their exile. To the extent Quenya was spoken in Middle-earth, it reflected these changes: At the end of polysyllabic words, all long vowels became short. This affected the long final vowels Quenya had reacquired after the pre-historic shortening of the long final vowels of primitive Elvish. It included - as the nominative plural ending of words in - (this long - being reduced from -ei at an older stage) as well as - in the genitive forms of nouns in -a (reduced from earlier -ao). Thus Book Quenya forms like lass leaves or ciry of a ship now became lassi, ciryo. This shortening of final vowels also resulted in the loss of distinct accusative forms that in Book Quenya were formed by making a short final vowel long (e.g. ciry as the accusative of cirya ship, but now ciry was reduced to cirya and merged with the nominative form).

CHANGES PRODUCING EXILIC QUENYA

55

Another change mentioned by Tolkien in the Plotz letter affects final -ai at the end of polysyllabic words. This diphthong first became a long - (with the same quality as the vowel of English cat, but longer). Then, like other long vowels at the end of polysyllabic words, it was shortened and turned into a regular final -. In this way adjectives in -a came to have plural forms in -, e.g. lint as the plural form of *linta swift, lint representing earlier *lintai. Adjectives in -a, that originally had plural forms in -ai, came to have plural forms in -i (obviously dissimilated from *-e). In Plotz, Tolkien also mentioned another late change: In polysyllabic words, a long vowel occurring before a final consonant might also be shortened. Some of these long vowels had occurred in a final syllable since ancient times, others had ended up in the final syllable by the loss of some final vowel. Shortened forms of names, such as Valinor (for *Valinr as a shorter form of Valinr) would also belong on this list:
Anr > Anar sun (SD:306) tern > teren slender (gajr > *ayr > *ayr >) r > ar sea (Mbelekro > *Mbelekr >) *Melkr > Melkor (*Erukhn > *Erukhna > *Erukhn > *Erun >) *Eruhn > Eruhin Child of Eru (but still Eruhn- before endings) *Valatr Vala-king > Valatar (but still Valatr- before endings) *Valinr (shortened form of Valinr) > Valinor

The Quenya word for Middle-earth is usually cited as Endor, a shortened form of Endr (both forms are mentioned in MR:126). Endr, possibly the intermediate form between Endor and Endr, occurs in the phrase Aran Endr King of Middle-earth (MR:121). Where a word occurs compounded or with endings, an originally long vowel might stay long because according to the normal stress rules it happens to receive the accent. The singular *Eruhn > *Eruhin thus retains the plural Eruhni with the long vowel intact. (But the plural of ar is ari rather than *ri; as we noted above, three-syllable words tend to move the stress to the first syllable and shorten the vowel in the second-to-last syllable.) Quenya as spoken in Valinor possessed voiceless l and r, spelt hl and hr. In Appendix E of LotR it is indicated7 that by the Third Age, hl was usually pronounced as [normal, voiced] l. It may be assumed that hr was likewise normally pronounced as a voiced r in Frodo's day, though there is no explicit statement to this effect. (However, it seems that the digraphs hl- and hr- persisted in the established spelling of Quenya.) Between i and another vowel, a glide consonant y could develop, breaking up the hiatus. Citing a word for shingle as sarni, Tolkien also listed sarniy as a parenthetical alternative (VT42:11). This glide consonant may also develop between a diphthong in -i and a following vowel, as when older waia, vaia envelope could also appear as waiya, vaiya (Etym, entry WAY). These examples suggest that it is difficult to pinpoint exactly when the glide consonant developed: was it before or after initial w- became v-, a change we tentatively assume to have occurred in the Second Age? As suggested by the example sarni(y), the presence of such glide consonants seems to be an optional feature of Exilic Quenya, and their representation in writing is apparently rather irregular. There is no trace of glide consonants in Namri in LotR; we have forms like lauri and namri rather than *lauriy, *namriy. On the other hand, Frodo speaking in tongues in Cirith Ungol employs the word aiya for hail!, this form representing older aia (so in Tolkien's Quenya translation of the Hail Marysee VT43:28).

56
The Evolution from Primitive Elvish to Quenya Lastly, certain assimilations seen to occur in contact may be listed; it is difficult to say at what stage these assimilations happened. Perhaps the combinations in question were assimilated already in Valinorean Quenya, but were treated likewise in younger compounds. The groups nl, nm, nr, ns are all assimilated to double consonants ll, mm, rr, ss:
Nmen West + lt flower = Nmellt Flower of the West elen star + macil sword = Elemmacil *Star-sword elen star + rna crowned = Elerrna Star-crowned elen star + sar stone = Elessar Star-stone (or Elf-stone)

The cluster ll may also arise from rl, e.g. Casalli as the partitive plural of Casar Dwarf (Casar + -li). Notice also that nm is now assimilated to mm instead of becoming nw, as in the prehistoric period.

EDITORIAL NOTES
1 2 3 See LotR:1096-97. Ibid, p.1096. Tolkien later changed his mind regarding samna diphthong. According to VT44:13, Tolkien rejected the base SAM in Etym, to be replaced by KAM, thus okamna in lieu of osamna. In any case, by the time Tolkien wrote Quendi and Eldar the Quenya words for diphthong became ohlon and ostime. Still there is another root SAM meaning mind, think, reflect, be aware (VT41:5), but as Helge himself wrote, these are hardly difficult to distinguish in context! Cf. p. 22 note 1. See LotR:1088, cf. 1096 That is, ibid, pp. 1096-97. That is, ibid, p. 1096. See also p. 10, note 2 above. That is, ibid, p. 1088 [cf. 1097, footnote]

4 5 6 7

APPENDIX: SOME SPECIAL CASES

I will briefly discuss certain words and phonological features that do not readily agree with the system sketched above. In Etym, Tolkien derived the name Melko (Melkor) from Mailik, an A-infixed variant of the extended base MIL-IK supplied with the masculine ending -. The name is apparently here meant to suggest *Greedy One. This seems like a strange ad hoc etymology. It is difficult to get from Mailik to Melko. The shortening of final - to -o is of course regular enough, but Tolkien here has the second i undergoing syncope in a word that does not have two identical short vowels following one another the environment normally triggering such syncope. Or are we to assume that the final element of the diphthong ai in the initial syllable constitutes the first of this pair of identical vowels? The intermediate form *Mailko is then supposed to become Melko, the diphthong ai collapsing into a monophthong e. This development is apparently meant to be triggered by the following consonant cluster; yet this does not agree with other material from Etym, cf. a word like aista- (entry GYAS) failing to become **esta-. In any case, Tolkien abandoned the etymology here proposed for the name Melko. In a post-LotR source, this alternative name of Melkor is said to mean simply Mighty One, derived from a stem melk- or melekhaving to do with force and strength (MR:350, 360). This mel(e)k- may be taken as the Classical Quenya form of the original mbelek- seen in the old form Mbelekro Melkor (WJ:402). So according to Tolkien's later ideas, the shorter name Melko is certainly meant to come from *Mbelek, and the troublesome evolution Mailik > Melko presupposed in Etym does not have to be considered part of the classical linguistic scenario.1 The same goes for malarauco as the Quenya word for Balrog. It is listed in Etym, entry RUK, and explicitly referred to a primitive form gwalarauk. However, the idea that initial gw- is to produce Quenya m- is blatantly contradicted by the entry GWAL (though there is actually a cross-reference to it in the same entry RUK!) From GWALTolkien derived words like nwalya to pain and nwalca cruelnot **malya and **malca, though the primitive forms are obviously meant to be *gwalj, *gwalk. We can be certain that the words nwalya and nwalca reflect Tolkien's final decision on how primitive initial *gw- is to come out in Quenya, for what is obviously another derivative from this stem occurs as the name of a Tengwa letter in LotR Appendix E2: nwalm torment (earlier ngwalme = walm). So why does gwalarauk produce malarauco instead of the expected form *nwalarauco in Etym? Comparison with earlier material reveals that malarauco as the High-elven word for Balrog is influenced by such early Qenya forms as Malkarauke (LT1:250). However, the m-forms can be ignored altogether, for Tolkien later decided that in Quenya, a Balrog is called a Valarauco (WJ:415; VT39:10). This is clearly meant to be derived from *balarauk, and the phonological evolution would be quite regular. The word palantr is somewhat puzzling. In the Plotz letter, Tolkien explicitly stated that a long vowel in the final syllable of a polysyllabic word became short; yet the long of palantr apparently persisted as late as in the Third Age. On the other hand, palantir with a short i is mentioned as the name of a Nmenorean king in the LotR appendices. In Letters:427, Tolkien refers palantr to a primitive form palantr or palantra; he marked the final -a with a diacritic indicating that it could be either long or short. (This

58
The Evolution from Primitive Elvish to Quenya primitive form must be seen as a hypothetical back-formation, since the palantri were made and presumably named by the Noldor after the end of the pre-historic period.) According to the rules set out above, palantr should have produced Quenya *palantra, whereas palantra should have yielded palantir, attested as the personal name of Tar-Palantir of Nmenor. (Palantr could also have produced palantir if the primitive form was recognized as a compound, so that the long final vowel was shortened at an early stage.) So what are we to make of the form palantr occurring in LotR? Perhaps we are to assume that since the Seeing Stones were so closely associated with ancient lore and were rarely mentioned otherwise, the word persisted in its older Valinorean form rather than undergoing the normal sound-changes? Of course, the long vowel may also have been re-introduced into the singular by analogy with the plural form palantri. Another example of a word with a long final vowel persisting in the final syllable is Erusn, translated children of God in RGEO:74. This form is doubly perplexing, for judging by the translation, it should be a plural form. Yet there is no plural ending. Is this a collective of some sort?3 The noun head is given in Etym as cr with stem cas-; the root is KAS. This makes little sense within the later phonological framework; it seems to be a form carried over from Tolkien's earliest Qenya (QL:45: notice that the K-stems are among the oldest parts of Etym, apparently never thoroughly revised). In Etym, Tolkien lets a final postvocalic -s become -r, as when the root TELES produces Quenya Teler Teler-elf. The root KAS, or rather a primitive form *kasa > Common Eldarin *ks, *kas-, might therefore produce cr via *cz. However, this would not have a stem-form cas-, e.g. plural **casi heads no more than the plural of Teler was **Telesi even within the Etym scenario (the plural Teleri occurs in the entry TELES). According to the system Tolkien employed in most of Etym, the word for head ought to be *cr with stem *car-. Later Tolkien apparently decided that a final original -s remained unchanged, though the development s > z > r does occur between vowels; thus we have olos dream, pl. olori for older olozi, in UT:396. If the noun olor of Etym (entry LOS) was later emended to olos, the noun cr head (KAS) must also become *cs, though the stem-form remains (*caz- >) *car-. A possible attestation of cas as a word for head occurs in the compound cas-raya tressure (literally *head-net), later becoming carra (VT42:12). Interestingly, the form cas-raya is not asterisked, and it is spelt with a c rather than k also in the primary source suggesting that this is a historical or attested Quenya form. Perhaps, then, cas-raya is meant to represent the form of the word in Valinorean Quenya of the early historical period, the classical form carra arising within recorded history. (The assimilation of medial sr to rr probably began already in the prehistoric period, but perhaps it continued to operate and was eventually applied even to more recent compounds.) According to Etym, the stem OL yielded the Quenya word holm odour. This must be derived from primitive *olm. So does initial *- produce Quenya h-? If so, this change must have occurred before the ancient initial combination g- was simplified to -, since the latter persisted in Valinorean Quenya (becoming normal n- in the Third Age). However, in a much later source, Tolkien lets primitive - remain unchanged in Quenya: the word alta radiance, glittering reflection he referred to primitive alat. So why does *olm produce Quenya holm instead of *olm? Some think holm is simply a misreading for *nolm in Tolkien's manuscript. *nolm could be the later pronunciation of *olm.4

APPENDIX: SOME SPECIAL CASES

59

The view presented above is that the shortening of the long final vowels of primitive Elvish took place at a relatively late stage (after Common Eldarin), whereas original short final -a, -e and -o disappeared during the Common Eldarin stage. Examples supporting this view are cited above. However, counterexamples can certainly be found. In some sources, Tolkien cites Common Eldarin words with short final vowels that survive unchanged into Quenya, as when the forms sloga and netere are said to be the origin of Quenya hla, nert (VT42:9,26). According to the system here reconstructed, these CE words would rather produce Quenya **hl, **neter. To reach the desired outcome in Quenya we would have to assume that the CE forms were *slog, *neter with long final vowels. We could assume that after original short -a, -e, -o were lost, but still during the Common Eldarin stage, the original long final vowels were shortened so that *slog, *neter turned into the forms sloga, netere mentioned by Tolkien. Some examples from Etym may also seem to support such a view, as when Tolkien mentioned daio as the primitive form of lo shade (LR:354 entry DAY). The final -o must in all likelihood have been a long - at some earlier stage, but here it may seem to have been shortened quite early, before initial d- had become l-. The latter change must in turn have occurred before initial and post-vocalic b, d, g became spirants v, (> z > r), , respectively, or primitive initial d- would have produced Quenya r- instead of l-. Yet this is totally contradicted by one attested example referred to above: Valinr as an early Quenya name of Valinor, with a long final vowel intact as late as after initial b- has become v- (WJ:413). There is hardly any way all the evidence can be worked into a single system. We must probably conclude that Tolkien in some cases failed to mark long final vowels as such. Of course, he may also have changed his mind (repeatedly!) about the precise order of certain phonological changes. In some sources, Tolkien seems to be toying with the idea that a double consonant is to be simplified before a long, accented vowel. Thus Elerrna *Star-crowned (assimilated from *Elenrna) appears as Elerna in LR:355. Likewise, the name of Tuor's ship Erm (WJ:352) is probably to be understood as Er + rm, i.e. *Sea-wing, rather than E + rm Universe-wing. However, Tolkien may have abandoned this idea. Elerrna is so spelt in the published Silmarillion, and I think this form is to be preferred. Finally, although Tolkien explicitly wrote PQ h survived in the dialects of Aman [i.e. Vanya-oldorin Quenya and Telerin] (WJ:365), there is evidence that this rule was not always followed. A case to consider is end and related derivatives of the base NED in Etym. It seemed that Tolkien changed his mind as to the root from which they were derived. In VT41:16 note 5 he proposed the base HEN or (extended form HENET). Thus we have end or ent (noun) and enda or entya (adj.). But compare Q Endor and T Hendor! In the same vein, in the manuscript version of Notes on re we have the root HOR urge, impel, move, from which is derived the word r heart, inner mind. But compare the impersonal verb or urge, impel, move with T. hore. There also exists a base HAN add to, increase, enhance, honour (esp. by gift) (VT43:14), from which we have derivatives han beyond and hantal thanksgiving. But there may be a relation between this root and ANA1 in Etym: to, towards from which we have the derivatives anna gift and anta to give. Therefore, I would not be surprised that aran and other derivatives of base ARA noble would have a related base *HAR (cf. AR in Etym.).5

60
The Evolution from Primitive Elvish to Quenya

EDITORIAL NOTES
1 Maybe so, but we can also consider the putative descendent form *Mailiko as a variant of Melko, a play on words of which the Eldar are pretty much capable. We have also the case of the name Sauron, which can be interpreted as the Abhorred (< thaurond- < root THAW) or the Foul One (< *thauzond- < TH, with a-infixion), or the name of the Vali Nessa (see p. 40 note 4). In Sindarin we have Belegr, the Sindarin cognate of Q Melkor but there is also the punning variant Belegurth Great Death. Although Tolkien explicitly wrote that the word Melko (and Melkor) is derived from a stem mel(e)k, it is reasonable to assume that there is a variant stem-form *MBELEK underlying this (Sindarin derivatives beleg, Belegr). And there is one more case that should bear consideration as well. In VT41:14 and 19 note 19 we encounter the name Ekyanro Sharp Flame (Sindarized Eignor) as a variant form of Aikanr(o) (Sindarized Aegnor); the first element of the former is suggested to by derived from stem EKE stab, prick with a sharp point, while that of the latter is from GAYA dread, awe. It is rather puzzling why Tolkien chose to present these forms so late in his lifeconsidering that the form Aikanro is attested in a contemporaneous source (PM:347)unless he intended it as a punning variant as well. However a better explanation for is is that this could also explain the existence of the Sindarin element aeg (as seen in the attested forms Aeglos and Hithaeglir) as an alteration of eig (< *ekj) when it was written explicitly that the descendant forms of the primitive adjectival form gajak were not preserved in Telerin and Sindarin (PM:363, note 45). If I am correct, then the genuine Sindarin forms would have been *eiglos, *Hitheiglir. That is, LotR:1096 But in MR:423 note 1 we encounter the form Erusni. Cf. Eruhni and Eruhn in VT39:34 note 5. I tend to agree with Helge; I have personal doubts as well that OL is the correct reading of the root form in Etym.But what if the actual stem is *HOL instead? This would have made better sense, for not only would have resulted in Quenya *holm, but also because primitive h disappeared altogether in Sindarin, this would have resulted in the descendant form l, and indeed it does. See also page 53, footnote.

2 3 4