APPENDICES

Based on all published material (and then some), I have set out most of the information about Quenya that I feel we can infer with a minimum of confidence. These Appendices to the course proper will supply some extra information, but primarily they will be devoted to discussin certain features of Quenya that are not so well attested or understood that I felt able to construct any exercises touchin on these matters. If more Quenya material is published in the future, I may also summari!e the new information here (unless or until I wor" it into the existin lessons, or write additional lessons to cover the fresh info). WHAT IS THE FUNCTION OF THE PARTITIVE PLURAL? In this course we have discussed three Quenya numbers# sin ular, plural, and dual. There is however also a fourth, called the partitive plural, i nored throu hout this course because I feel its function is not well understood. $ence I am also unable to uarantee that Tol"ien would not have used the partitive plural where I have used the %normal% plural in some of the exercises of this course. Before discussin the possible function of the partitive plural, we will discuss how it is formed. The basic endin is &li, derived from a root LI %many% which is also the source of the noun li %people%. The 'lot! (etter points to lasseli, ciryali as the partitive plural forms of lass %leaf% and cirya %ship%. It is less than clear how &li can be added to nouns endin in a consonant, for this would normally result in impossible consonant clusters (only nouns endin in &l can receive the endin &li with no further ado, since Quenya does permit double ll). In one attested example, involvin the partitive plural of Casar %dwarf% (based on )warvish Khazd and an alternative to the native word Nauco), we see assimilation# instead of the impossible form **Casarli we have Casalli. The Etymologies li"ewise points to Telelli, not **Telerli, as the partitive plural of Teler (an +lf of the Third ,lan)- see ./#011, (2#034. (The double asteris" ** is here used to mar" wron forms.) 5ouns endin in &s or &n may also assimilate this consonant to l before the endin &li, so perhaps the partitive plural of nouns li"e elen %star% or olos %dream% would be elelli, ololli. The behaviour of nouns endin in & , li"e na %thin %, must remain a mystery for now. +ither the impossible form **na li would have its consonants swapped around, producin nal i, or we must probably slip in some connectin vowel 6 perhaps resultin in a form li"e na eli (with the same extra vowel &e& as in Elen!ilenna %to +lendil%, '7#894). As for the various case forms as set out in the 'lot! (etter, the archaic %Boo" Quenya% is to have the lon endin &l" in the nominative and the accusative ali"e, but in later Quenya, this is shortened to &li as in the examples above. The endin &l"& with a lon vowel is however still used before the endin s &nen for instrumental and &#a for possessive, so in these two cases, the partitive plural form ciryali appears as ciryal"nen and ciryal"#a respectively (and the

lon vowel of course receives the accent). The enitive simply adds the endin &on as in the normal plural, hence ciryalion. The dative adds the simplest dative endin &n, hence ciryalin. In the locative, allative and ablative, it is optional whether one uses the simplest endin s &ss, &nna, &llo or their plural forms &ssen, &nnar or &llon (alternatively &llor). The endin &li& already indicates that the word is plural, so whether or not a second plural indicator follows at the end of the word is not important. $ence locative ciryaliss : ciryalissen, allative ciryalinna : ciryalinnar, ablative ciryalillo : ciryalillon (or ciryalillor). In Namri in (ot2, Tol"ien used $al%alinnar as the partitive plural allative of the noun $al%a %(foamin ) wave%, so writers who want to use a strictly (ot2&style form of Quenya may opt for the forms with double plural mar"in . (istin the forms is easy- it is rather more difficult to tell precisely what their function is. If ciryar is simply %ships%, how does the alternative plural form ciryali differ in meanin ; In +n lish translation, Tol"ien renders the %partitive plural% forms in &li as normal +n lish plural forms (in &s)# The allative phrase i $al%alinnar in Namri is translated %upon the foamin waves%. $owever, in his interlinear translation of Namri in 2<+=#>?, Tol"ien bro"e this form of $al%a down as $al%a&li&nnar and indicated that the middle element means %many% 6 which as we have seen is also the meanin of the basic stem LI ((2#0>3). @urthermore, an +nt on one occasion used the word aurelil'%a, which in (ot2 Appendix @ is said to mean %@orest&many&shadowed%. Thou h this is not enuine Quenya as such, but merely %fra ments of +lf&speech strun to ether in +nt&fashion%, Tol"ien is a ain seen to imply that the element &li connotes %many%. Ao inevitably and not without reason, many researchers have concluded that the forms in &li are examples of a so&called multiple plural. This plural form was thou ht to imply that there are many of the thin in Buestion# .hereas ciryar is simply %ships% (few or many, but at least more than one), ciryali would imply %many ships%. The example i $al%alinnar %upon the foamin waves% in Namri would a ree well with this interpretation# <aladriel is sin in about the waves of the wide ocean between herself and Carda 6 obviously a reat multitude of waves. $owever, I fear that this interpretation of the plural forms in &li is too simplistic- at least this can hardly be the whole story. Treebeard reeted ,eleborn and <aladriel as a #ani%ar( #ani%)lion nos ari, which reetin Tol"ien translated as %o fair ones be etters of fair ones% (A)#?0). Vani%)lion %of fair ones% is the partitive plural enitive of #ani%a %fair (one)%. A ain, Tol"ien translates a partitive plural form as a normal +n lish plural, and there is nothin to su est that the meanin %of many fair ones% is intended. (Dnless Treebeard exa erates to be polite, this would not a ree very well with the facts either# ,eleborn and <aladriel were the %be etters% of one "nown child, their

dau hter ,elebrEan. +ven if we throw in their randdau hter Arwen as well, there are still only two %fair ones%- this can hardly count as %many%.) There is also an example of a partitive plural (this time in the locative case) in the Mar irya poem, but it is not very helpful# *an cenu#a r)cina cirya on!oliss %orn, which Buestion Tol"ien rendered %who shall heed a bro"en ship on the dar" roc"sF;G% (7,#HHH, cf. H4I, HH9). Ao here we have on!oliss %orn : %on FtheG dar" roc"s%- once a ain, a Quenya plural in &li is translated as a normal +n lish plural in &s. .hile there is nothin that would preclude the possibility that the ship is described as bein stranded on many %dar" roc"s%, there is nothin that would confirm this interpretation, either. +arly material (far predatin the (ot2) provides yet more examples of plurals in &li. In the early versions of the !ilmarillion narratives, the clan of the 5oldor are referred to as the Nol!oli (e. . (T4#H4). $ere the endin &li may seem to be used as a "ind of % eneric plural%, referrin to the entire %race% of 5oldorin +lves. Aimilarly, in the entry "ELE! in the Etymologies ((2#034), Tol"ien may seem to virtually eBuate the form Telelli with the compound Telelli %Teler&fol"% (Teler %Telerin +lf% J li %people%- notice the re ular assimilation rl K ll). $owever, when discussin various forms of Casar %)warf% in a much later (post&(ot2) source, Tol"ien seems to clearly distin uish between %the partitive plural% Casalli and %the race&name% Casalli (the latter evidently : %)warf&people%, Casar J li). Aee ./#011. In the later narratives, Tol"ien also abandons the form Nol!oli and consistently refers to this people as the Nol!or 6 a %normal% plural in &r, even thou h the reference is clearly to the entire clan of 5oldorin +lves as a %people%. The form Nol!oli was certainly not obsolete as such, but perhaps its function was redefined. In one passa e in Letters, Tol"ien ma"es some remar"s that throw at least some li ht on the various plural formations, but he uses Aindarin examples# The +ldarin lan ua es distin uish in form and use between a LpartitiveL or LparticularL plural, and the eneral or total plural. Thus yrch Lorcs, some orcs...L ... the =rcs, as a race, or the whole of a roup previously mentioned would have been orchoth. ((etters#4?1) If yrch %=rcs% is an example of a Aindarin %partitive% or %particular% plural, it must correspond to a Quenya plural in &li (thou h historically, yrch is rather descended from a plural in &i, which plural endin still survives in Quenya). .e must then eBuate the %normal% Quenya plural in &i or &r with the % eneral or total plural%. Accordin to what Tol"ien says, this plural would often be used of entire races (or of distinct groups previously identified). Indeed we see forms li"e Valar, +uen!i, El!ar referrin to the relevant %races% (does this obsolete the notion that the plural in &li can also have this meanin , and is this why Tol"ien replaced Nol!oli with Nol!or in his narratives;) =f course, these plural forms can also have a more limited reference, pointin to some particular roup rather

than the entire %species% of the thin in Buestion. Treebeard addresses <aladriel and ,eleborn as a #ani%ar %o fair ones%- they are obviously Must a couple of fair (beautiful) persons, not by themselves constitutin the entire %race% of fair people in the world. 'erhaps the system wor"s somethin li"e this# If you refer to a roup usin the %normal% plural in &r or &i and do not insert the definite article #efore the plural noun, the noun may often have eneric reference unless the context clearly indicates otherwise. ,f. a Tol"ien&made example li"e Valar #alu#ar, %the will of FallG the Calar F<ods, 'owersG will be done% (./#898). Valar is here a % eneral or total% plural- it does not refer to %some% ods or 'owers, but to the entire %race% of Calar. If you want to introduce some Calar as opposed to the entire race of ods, it would perhaps be appropriate to use the partitive plural# Valali. (,f. Tol"ienLs implication that %some orcs% is a possible renderin of the Aindarin partitive plural yrch.) The very term partitive plural would then su est that we are dealin with a sub& roup, a part of the total roup of Calar in the world. But once these Valali have been established as %a roup previously mentioned%, we would switch bac" to the %total% plural, that now refers to the total of the aforementioned sub& roup. 5ow it would be appropriate to spea" of this sub& roup as i Valar (with the definite article# the particular ods we are considerin here). .hen Treebeard addresses <aladriel and ,eleborn as a #ani%ar %o fair ones%, he naturally uses the %particular% plural since he is addressin two particular %fair% people. But when he oes on to describe them as #ani%)lion nos ari %be etters of fair ones%, he switches to the %partitive% plural to ma"e it clear that <aladriel and ,eleborn are the ( rand)parents of some %fair ones% as opposed to all the fair ones in the world. ('erhaps TreebeardLs wordin could also be ta"en to mean that some, but not all of <aladrielLs children were %fair% or beautiful. $owever, the context, as well as eneral courtesy, would indicate otherwise.) As for the ship that is stranded on!oliss, it is simply described as bein stranded on some roc"s, certain roc"s, a num#er of roc"s. Ao in short, it may be that Quenya would typically use the plurals in &li where +n lish would have %some% J a plural noun. Net the older theory of the %multiple plural% may not be entirely wron - the fact remains that the root LI ori inally si nifies %many%. 'erhaps it is especially when plurals in &li are combined with the definite article that these forms may carry the idea of a reat multitude, as in the Namri phrase i $al%alinnar %upon the F;vast ocean ofG foamin waves%. But enerally, it may be Must as wise to use a separate adMective to indicate %many% (in this course we used the adMective ri%,a pl. ri%,, by Tol"ien lossed %numerous%, to express this meanin ). Aince we have so few examples, these theories must remain tentative. As I said, I certainly cannot rule out the possibility that Tol"ien would sometimes have used plurals in &li where I have used %normal% plurals in the exercises of

this course. 7uch less can I tell whether he would have reMected the alternative wordin as downri ht %wron % or un rammatical Quenya. THE APPLICATION OF CASE ENDIN-S INCLUDIN- T As we recall, there are two Quenya endin s denotin dual number# &u and & (e. . Al!u %two trees%, cirya %two ships%). .hich endin is used depends on the shape of the noun it is added to. In the 'lot! (etter, Tol"ien listed various case endin s also includin the dual element & &# enitive & o, dative &n , allative &n a, ablative &l o, locative & s, instrumental &n en. =bviously, these are mere variations of the simplest case endin s, normally associated with the sin ular# &o, &n, &nna, &llo, &ss, &nen. The correspondin dual endin s are derived simply by sBuee!in in a & & and (where necessary) adaptin the result to fit Quenya phonolo y. Ao the Buestion is# since & is obviously the dual mar"er here, are case endin s includin & & to be attached only to nouns with nominative dual forms in & , li"e cirya ; 7aybe nouns that in the nominative receive the other dual mar"er, &u, should not ta"e case endin s includin & &; This is a Buestion I briefly discussed in many of the lessons, and as I indicated, there is no certain answer available. Net we may form a plausible theory if we can fi ure out how the case endin s includin & & ori inally arose in the lan ua e. 'erhaps Tol"ien ima ined that initially, the case endin s were simply added to the simplest dual form in & . Thus, startin from the simple nominative cirya %two ships, a couple of ships%# cirya cirya cirya cirya cirya cirya J &o for enitive : cirya o J &n for dative : cirya n J &nna for allative : cirya nna, simplified to cirya na J &llo for ablative : cirya llo, simplified to cirya lo J &ss for locative : cirya ss, simplified to cirya s J &nen for instrumental : cirya nen

$owever, the roup n came to be disli"ed, so the consonants underwent metathesis- that is, they were swapped around to produce n instead. Thus the dative cirya n, the allative cirya na and the instrumental cirya nen turn into the actual forms listed in 'lot!# ciryan , ciryan a, ciryan en. (i"ewise, l is swapped around to l , so that the ablative chan es from cirya lo to the attested 'lot! form ciryal o. =nly the enitive cirya o and the locative cirya s persisted as they were, with no metathesis 6 ivin away that ori inally, the case endin s were suffixed directly to the simplest dual form in & .

If this is so, we have every reason to assume that the same case endin s were suffixed to dual forms in &u as well, for instance li"e this (usin Al!u %Two Trees% as our standard example)# Al!u J &o for enitive : Al!uo Al!u J &n for dative : Al!un Al!u J &nna for allative : Al!unna Al!u J &llo for ablative : Al!ullo Al!u J &ss for locative : Al!uss Al!u J &nen for instrumental : Al!unen These forms would under o no further chan es, since they are all acceptable Quenya as far as phonolo y oes. By this theory, the dual case endin s &n , &n a, &l o, & s and &n en should only be suffixed to nouns with nominative dual forms in & . 5ouns with nominative dual forms in &u would merely add the simplest case endin s &o, &n, &nna, &llo, &ss, &nen. The sole uncertainty would relate to the dative form. It could be Al!un as su ested before, but as we demonstrated in (esson Thirteen, Tol"ien used the lon er endin &en in one attested example datin from the period when &(e)n was still the enitive rather than the dative endin # Veruen as the enitive of the dual form #eru %married pair, spouses% (+tym,. entry LE$, cf. %E!). If this formation survived as such, irrespective of Tol"ienLs re&definin of this case endin , the dative of Al!u should be Al!uen. As for the possessive&adMectival case form, no dual endin s are attested anywhere. .e have theori!ed that it should have the endin & .a in the case of dual forms in & (cirya .a %of a couple of ships%), but if the theory set out above is correct, the endin should simply be &#a in the case of dual forms in &u (Al!u#a %of FtheG Two Trees%). INFINITIVES WITH PREFI/ED A& The Mar irya poem includes two examples of a special infinitive that is mar"ed with the prefix a&. (Tol"ien first wrote na&, then chan ed it to a&, perhaps su estin that this was a rammatical feature that arose spontaneously in his mind as he wor"ed out the later version of this poem.) The relevant couple of consecutive lines o li"e this# *an cenu#a lu%,or a0os a %.ho shall see FtheG clouds ather,% *enel ac1na... %the heavens bendin ...F;G% The verbs involved are the A&stems 0os a& %to ather% and c1na& %to bend%. .hat is the function of the a&prefix here occurrin ; Tol"ien provided a brief, rather obscure note about it# %.hen the bare stem of the verb is used (as after LseeL or LhearL) as infinitive Fa&G is prefixed if the noun is the obMect not the subMect% (7,#HH0). There are at least two possible interpretations of this, but the %traditional% interpretation 6 reflected in many

post&Tol"ien texts 6 oes li"e this# @ollowin sense&verbs (li"e %see%, %watch%, %hear% etc.) we will typically find some noun that is the o#&ect of the sense&verb in Buestion. In the examples above, lu%,or %clouds% and *enel %the heavens% are the o#&ects (not the subMectsO) of the verb cenu#a, %shall see%. 5onetheless, these verbs are also the lo ical su#&ects of the verbs %to ather% and %to bend%# The clouds ather, the heavens bend. Ao to express what an obMect is itself doing while it is the subMect of another verb, you use an infinitive with a prefixed a&. In other words, %an cenu#a lu%,or a0os a is the Quenya way of sayin %who shall see FtheG clouds as they ather%. But unli"e % ather% in this +n lish circumlocution, a0os a remains an infinitive, as is clearly seen by the fact that it does not receive the plural endin &r even thou h its lo ical subMect (lu%,or %clouds%) is plural. .e may wonder how the prefix a& would be added to a verbal stem be innin in a vowel, especially in a&. As su ested by 5ancy 7artsch in her %asic 'uenya, it may sometimes be best to slip in a hyphen, for instance in a sentence li"e cennen i n"s a&an a F;G i sel!on 2ar%a, %I saw the woman ive the boy a boo"%. It is also uncertain how the prefix a& would be added to primary verbs. A verb li"e (say) %a & %to eat% would probably include the endin & in its %stem% form, Must as it does when it functions as an infinitive. =f course we may prefix a& and build a sentence li"e irnen i Naucor a%a : %I watched the )warves eat% (: %I watched the )warves as they ate%). $owever, a form li"e a%a would be accented on the first syllable, on the prefix itself. 'erhaps we should i nore the normal stress rules and let the accent fall on the syllable followin the prefix (the spellin a&%a could be employed to su est this); Actually, the Mar irya poem contains a relevant example involvin the verb cir& %to sail%, but in this example, the prefix a& is not used at all. ,orrectin some li"ely misreadin s (a in Tol"ienLs manuscript thrice bein misread as e), this line would o# *an cenu#a $)na cirya %3 i%a 0ras allo c"ra : %.ho shall see a white ship sail from the last shoreF;G% (Tol"ienLs poetic translation in 7,#H48 reads %who shall see a white ship leave the last shore% 6 but it is clear what the literal meanin is.) The overall construction is very similar to the examples cited above- the %white ship% is the o#&ect of the verb cenu#a : %shall see%, but it is also the lo ical su#&ect of the verb cir& %to sail%. The latter is here simply constructed as a %continuative% stem, similar in form to the %present% or %continuous% tense# c"ra. Ao we must assume that %I watched the )warves eat% could similarly be expressed as irnen i Naucor %) a, and I thin" I would prefer this construction to the uncertain a%a &form. =f course, we may then as" why a simple %continuative% stem cannot replace the infinitive with prefixed a& everywhere. The verbs 0os a& %to ather% and c1na& %to collect% would presumably produce the forms 0os a and c1na. .hy, then, cannot %who shall see the clouds ather, the heavens bendin ;% be expressed as %an cenu#a lu%,or 0os a( *enel c1na; .hy did Tol"ien use

the forms a0os a, ac1na instead; =f course, for all we "now, both constructions could be eBually possible, and Tol"ien simply pic"ed the one that came into his head first, or the one that fit the meter of his poem better. $owever, %<ildor In lorion% has su ested a new interpretation of the function of the a& prefix that would allow us to explain all the attested forms. .e have already Buoted Tol"ienLs note that %when the bare stem of the verb is used (as after LseeL or LhearL) as infinitive Fa&G is prefixed if the noun is the obMect not the subMect% (7,#HH0). )oes he mean# when the %noun% (that is, the obMect of the main verb of the sentence) is the lo ical obMect 6 not the subMect 6 of the infinitive itself; If so, it is obvious why the a&prefix is not used in a sentence li"e %who shall see a white ship sail...%, for the ship is the lo ical su#&ect of the verb %sail%, not the obMect. This interpretation would mean that the verbs 0os a& % ather% and c1na& %bend% are actually transitive in the examples where the a& prefix occurs, not intransitive as in the +n lish translation# %.ho shall see the clouds ather, the heavens bendin ;% In +n lish, it is Must the clouds themselves that % ather% or assemble, and the heavens themselves that %bend%- they donLt % ather% or %bend% somethin else. Net 0os a& is also lossed %collect%, which is clearly transitive. =f course, this verb could be both transitive and intransitive, Must li"e at least one other verb in & a (or a& : transitive %raise% or intransitive %rise%). But if 0os a& is ta"en as meanin % ather% in the transitive sense of %collect%, and the prefix a& indicates that the %noun% next to the infinitive is %the obMect not the subMect% of this verb, then %an cenu#a lu%,or a0os a4?5 actually means %who shall see the clouds #eing gathered(% Aimilarly, %an cenu#a666*enel ac1na would actually mean %who shall see...the heavens #eing #ent(% By this interpretation of the a& prefix, our home&made example irnen i Naucor a%a would not mean %I watched the )warves eat%, but rather %I watched the )warves bein eaten%O The prefix a& would indicate that the noun in front of the infinitive is actually to be ta"en as %the obMect not the subMect% of the eatin concerned. 'erhaps we shall never "now with certainty which of these two interpretations is correct. It may very well be that the prefix a& (variant na&) was spontaneously invented when Tol"ien wor"ed out the latest version of the Mar irya poem, and that it occurs nowhere else in his notes. ELISION OF FINAL VOWELS The reetin elen s"la l1%enn7 o%en iel#o %a star shines on the hour of our meetin % exemplifies a freBuent, but not obli atory feature of Quenya phonolo y# .hen a word ends in a vowel and the next word be ins in one, the former vowel may drop out. Thus the final &a of l1%enna %on FtheG hour% is omitted before the initial o& of o%en iel#o %of our meetin %. In +lendilLs =ath occur the words enn7 A%,ar&%e a, %until the end of the world%, the preposition enna %until% bein reduced to ennL before the initial a& of A%,ar&

%e a %.orld&end : the end of the world%. =ccasionally, this phenomenon may even result in new, seemin ly unitary words, as when the same preposition enna %until% and the noun oio %endless period% are contracted# ennoio (for enn7 oio) %forever% (DT#09I, 04?). .hen a vowel drops out li"e this, sounds and combinations not normally permitted finally may occur in this position# li"e the double &nn in l1%enn7, or the %final% &% in )%7 e ele0 a %deliver us% (CT80#4H, where the full form )% e ele0 a is also Buoted). 5aturally, such omission of final vowels is especially common when one word ends in a vowel that is similar or identical to the initial vowel of the next word (li"e the omission of &a before o& and a& in the examples above). In )*riel+s !ong as printed in (2#?H, there is a dot under some final vowels, which may be ta"en as a si n that they are to be elided. 2eplacin the dot with underlinin and not otherwise alterin the ori inal spellin of this %Qenya% text, we may Buote the relevant phrases li"e this# 4) Ilu Il1#a ar en 8)re el!ain %IlPvatar made the world for +lves%- here the final &e of 8)re %made% is apparently omitted before an identical vowel- H) "re il9a y3#a n' ina %when all is counted% and "re Anarinya 9elu#a %when my sun shall fail%- here we have omission of final &e before i and a, respectively- 0) eny)re ar i yel %in that day beyond the end%, which if we interpret Tol"ienLs intentions correctly is a uniBue example of a final vowel bein elided before an initial consonant in the next word ( ar). $owever, the last example has been Buestioned- the printed version may be in error when placin a dot under the final &e of eny)re. (Aee Tol"(an messa e 40.?I by )avid Aalo- than"s to )ie o Ae uE for brin in this to my attention.) .hatever the case, such omission of final vowels is clearly not necessary to produce correct Quenya, which is why we i nored it in the exercises of this course. The phrase l1%enn7 o%en iel#o is twice attested in its full form l1%enna o%en iel#o (./#0>?, (etters#8H8). =ur small corpus also contains examples of a final vowel persistin even where the next word be ins in an identical vowel 6 as in the sentence aur en ulu#a %day shall come a ain%, cited in the !ilmarillion, chapter H9. 'resumably this could also be shortened (aur7 en ulu#a), but not necessarily so. +lision of final vowels would probably be most common in the spo"en lan ua e, and in poetry it may also be useful to be able to et rid of a syllable where the poetic meter demands it. HISTORICALL: ;USTIFIED FOR*S( OR ANALO-ICAL LEVELIN-? As discussed in the Quenya lessons, seemin ly %irre ular% features of the lan ua e are very often Mustified by the lon historical evolution Tol"ien envisioned. @or instance, when the noun alan %floor% has the plural form ala%i instead of alani, this is because the ori inal 'rimitive +lvish base had the form ",L,M# As the distinctive features of Quenya phonolo y evolved, final &m was no lon er tolerated and was altered to the closest %permissible% sound# &n. $ence older talam appears as alan when the word occurs without endin s. But when

endin s are added so that a vowel follows, the ori inal &m was not final and therefore did not have to be chan ed. $ence the plural form ala%i %floors%. Net there could also have been another outcome# By analo y with such pairs as aran Q arani %"in Q "in s%, elen Q eleni %star Q stars% and many others, the pair alan Q ala%i %floor Q floors% could have succumbed to so&called analogical leveling. Apea"ers mi ht simply have fitted alan into the simplest pattern, so that its plural would have become alani instead. In this case, Tol"ien ima ined the historically Mustified form to have persisted. Net analo ical forms are not unheard of in his lan ua es (see for instance the entry $-ILIK in the Etymologies). The Quenyaist has to face the problem that it is impossible to predict with confidence to what extent Tol"ien meant analo ical forms to have replaced the historically Mustified ones. The silent assumption underlyin some of the forms presented in this course is that analo ical levelin had one Buite far, conveniently wipin out many of the extra complexities and %irre ularities% which students would otherwise have to deal with. Net we cannot be Buite sure that this is how Tol"ien himself envisioned his Quenya. Aome of the potential complications may be briefly addressed here. =ne problem has to do with the augment, the prefixed stem&vowel added to the verbal stem in the perfect tense# ul& %come%, but u 1li %has come%. 'refixin a vowel in such a way chan es the phonolo ical environment where the followin consonant occurs. If we start loo"in into the evolution of Quenya from 'rimitive +lvish, this could in some cases mean that the consonant itself would chan e. Ta"e, for instance, a verb li"e lan a& %fall%. 7ost writers have assumed that its perfect form, %has fallen%, would be alan i. A minute and a half into the credits of 'eter /ac"sonLs )ellowship of the .ing, you can even hear +nya sin %orni alan i, %dar"ness has fallen%. The form alan i is repeatedly used in this course as well. Net one could plausibly ar ue that the perfect tense of lan a& should actually be aran iO .hy; Because lan a& is derived from a stem /,"&, or specifically its nasal&infixed variant /,N"&. In early Quenya, the initial d& of 'rimitive +lvish turned into l& (./#0I0- exceptionally d& could also become n&). $ence we have lan a& %fall% from primitive dant&. $owever, primitive d developed in a Buite different way where it was not initial. @ollowin a vowel it turned into < and later mer ed with r- for instance, we have discussed how %ir %into% is evidently meant to come from primitive mi0da %in&to% (see the note in (esson @ourteen). Ao if the perfect au ment was prefixed to the stem dant& already before the initial d& turned into l&, older adant& would re ularly evolve into a<an & and then aran &, so that %has fallen% would be aran i rather than alan iO If this is correct, otherwise identical verbs would sometimes remain distinct in the perfect tense. In Quenya there are two verbs la#&, one meanin %lic"% and the other %yield, allow, rant%. The former is derived from a stem

L,%, so that the initial l& is ori inal- the latter comes from a stem /,% and thus mer ed with the other verb only when initial d& turned into l&. La#& %lic"% from L,% would have the strai htforward perfect form al)#i, whereas la#& %allow% from /,% would have the perfect form ar)#i (older a<)#i). By this system one would have to "now, in each case, whether the initial l& of any verb is ori inal or derived from older d& before one could construct the perfect formO But it does not stop here- we have Must opened up a 'andoraLs Box of potential extra complications, so that Quenya suddenly loo"s more li"e the %extremely difficult% lan ua e Tol"ien actually insisted it was ((etters#890). (i"e many African lan ua es of our own a e, 'rimitive +lvish was not averse to the initial nasali!ed stops m#&, ng& and nd&- Aouth African president Thabo 7be"i would presumably be more able to pronounce ancient +lvish words li"e m#undu %snout% than most .esteners are. In Quenya, the ancient initial m#& had been simplified to %&# m#undu came out as Q. %un!o (+tym, entry M%1/). The ori inal initial nd& of primitive +lvish was li"ewise simplified to n&, as when Quenya nulla %dus"y, obscure% is said to come from older ndulla (+tym, entry N/1L). =ri inal initial ng& first became =&, to use Tol"ienLs freBuent spellin of the sound of ng as in ing (pronounced without any distinct g, that is). @or instance, we have >ol!o from primitive ngolod2 (or technically 3golod2). By the Third A e, initial =& had come to be pronounced Must li"e normal n&# $ence the spellin Nol!or rather than >ol!or in (ot2 (but apparently, the distinction between =& and n& was upheld in Ten war writin ). To summari!e# ori inal m#&, nd&, 3g& became %&, n&, =&, respectively, and in spo"en Quenya, =& later mer ed with n&. %ut 6 this chan e only happened when the old initial combinations m#&, nd&, 3g& occurred at the #eginning of wordsO @ollowin a vowel in the middle of words, these combinations survived as they were. Thus in Quenya the root N/IL yieldin words for %friend, friendship, devotion% persists unchan ed in a compound li"e El!an!il %+lf&friend% (./#84H), thou h nd& has been simplified to n& in a related word li"e nil% %friendship%. ,ompare A)#H84, where Tol"ienLs character (owdham discusses these phenomena (referrin to Quenya as %Avallonian%). .hat is relevant for our purpose is of course that if a verb in %& or n& is derived from a stem in m#& or nd&Q3g&, one could ar ue that these combinations would survive followin the au ment that is prefixed in the perfect tense. Cerbs li"e na%,a& %to hammer% (stem N/,M&), nan!a& %to harp% (older =an!a&, stem 45,N) and %ar& %dwell% (stem M%,.) may then conceivably appear as an!a%,i, an?an!i, a%,)ri in the perfect tense. =f course, verbs in n& and %& that had Must been simple n& and %& all alon would not behave in this way. If a simple verbal stem li"e nac& %bite% is all one has to o on, it would be Buite unpredictable whether the perfect tense is to be an)ci (ori inal stem N,K) or an!)ci (as if the stem had been N/,K& instead). Cerbs with initial 0& would also be problematic. Aometimes Quenya 0& is derived from the 'rimitive +lvish aspirate h& (see below), which would evolve

into 0 also followin a vowel, but sometimes 0& is meant to come from primitive s &, as when the stem !K,"& yields the Quenya verb 0a & %brea" asunder%. 'rimitive s & becomes 0& only at the be innin of words. @ollowin a vowel, this cluster is sometimes unchan ed and sometimes has its consonants swapped around to produce 8s (or in re ulari!ed spellin @)- Tol"ien is Buite inconsistent in this matter. In +tym, entry MI!K&, he lists the adMective %i8sa (or, %i@a) %wet%- the entry&head would su est that the 8s of the Quenya word here comes from older s . In later sources Tol"ien has the cluster s8 (sc) survivin unchan ed in Quenya, as in rusco %fox% ('7#0I0, CT84#49). Ao one could ar ue that the perfect tense of a verb li"e 0a & %brea" asunder% should be, not a0) i, but either a@) i or asc) i, since the ori inal root is !K,"& and the perfect should conceivably descend from as t&. A verb asca & actually appears in one Quenya text (in the nasal&infixed past tense# ascan , evidently meanin %bro"e asunder% 6 see A)#049, where the spellin used is %as"ante%). In ascan , the prefixed stem&vowel is apparently used as an intensifier and is not really the perfect&tense au ment as such, but perhaps this au ment would have a similar effect on the rest of the word. And what about verbs in s&; 'rimitive s& was unchan ed at the be innin of words, but between vowels it was normally voiced to < and then mer ed with r. Ao perhaps the perfect tense of a verb li"e sal2a& %sip% should be, not asal2i, but rather aral2i for older a<al2i; =n the other hand, if s& represents older & (more or less li"e the %th% of +n lish thing), we would see &s& also between vowels# The perfect of sin a& %fade% would be isin i, since the ori inal root is "-IN& and the older Quenya forms would have been in a& with perfect iin i (the correspondin spellin is said to persist in Ten war ortho raphy). But then, Quenya initial s& can also come from the primitive initial cluster st&, which would be preserved unchanged between vowels. $owever, currently no Quenya verb derived from a stem in st& is "nown. 'erhaps we should be rateful. Then we have the primitive combination sy&, which produces Quenya 0y& when it occurs initially, but followin a vowel, sy apparently becomes ry (as when Tol"ien in +tym derives 2irya %syrup% from a stem $I!- the primitive form of the word must have been pisy). 7aybe the perfect tense of a verb li"e 0yar& %cleave% (stem !6,/&) is ary)ri rather than a0y)ri, then; And what about initial 0l&; It is derived from older sl&, a combination which would probably turn into &ll& between vowels. (At least primitive sr&, which at the be innin of words produces Quenya 0r&, is seen to become &rr& between vowels# Tol"ien derived Quenya %irroan. %incarnate% from primitive mi0 srawanwe, 72#0I9.) Ao maybe the perfect tense of 0lar& %hear% is not really a0l)ri, but rather all)ri;
5=T+# Aince I wrote the above, a note by Tol"ien has emer ed where he states that %in compounds with clearly perceived prefixes (or between clearly analy!ed and separate other elements) these Fthat is, 0l, 0r, 0., 0yG remain Funchan edG% (CT81#H3). $urray. As lon as the au ment used in the perfect tense is perceived as %clearly analy!ed and separate%, we can indeed o for a0l)ri rather than all)ri etc.

In short# .ithin Tol"ienLs eneral system there is room for many extra complications here, if one wants to draw the full conseBuences of the underlyin phonolo ical evolution that he ima ined. The au ment, the prefixed stem&vowel occurrin in the perfect tense, does seem to have been prefixed so early that the followin consonants still had not assumed the shape initial consonants would have in Quenya. (In ./#0>>, Tol"ien cites some %prehistoric% perfects already featurin the au ment.) To return to our first example, the d& of the stem /,7N8"& %fall% had not yet become l& as in lan a, and then the perfect %has fallen% could plausibly be (a<an i K) aran i because ori inal d became < and then r followin a vowel. Ao should we et in touch with +nya and tell her that she must record her son all over a ain, with %orni aran i instead of alan i; .e cannot tell. In this and all the other phonolo ical cases Must listed, it is simply impossible to predict to what extent Tol"ien meant the %historically Mustified% forms to have succumbed to analo ical levelin 6 if such levelin occurred at all in au mented verbs. Indeed, at present we cannot even "now whether Tol"ien ever considered these thin s. <iven the history Tol"ien ima ined for Quenya 6 that in 7iddle&earth it was nobodyLs mother&ton ue, but merely an ancient ritual lan ua e 6 it seems Buite plausible that the rammar of +xilic Quenya would tend to become somewhat simplified. Lan a& as a verb corresponds to the noun lan a, compounded in lasselan a %leaf&fall, autumn% (this word occurs in Appendix ) of (ot2, ma"in this form about as canonical as any Quenya word can be). If the compound was old enou h, this %should% have been **lasseran a instead, since the ori inal d of dant& is here intervocalic. Another possibility that we considered above was that verbs derived from ori inal stems in m#&, nd& and 3g& would still preserve these combinations followin the perfect au ment, so that (say) %ar& %dwell% from the stem M%,. should have the perfect form a%,)ri instead of a%)ri. It may be noted that in the Etymologies, Tol"ien derived the Quenya word for elephant 6 an!a%un!a 6 from primitive andam#und %lon &snouted% (see entry M%1/). This Quenya form could Must as well have been **an!a%,un!a, for ori inal m# could well survive in this position. Net the second element %&snouted% may here seem to have been altered from &%,un!a to &%un!a by analo y with such forms as %un!o %nose% (itself descended from primitive m#undu, initial m#& re ularly producin Quenya %&). =f course, such compounds may not tell us how au mented perfect forms would behave- the latter would probably be perceived as unitary words, whereas compounds are more obviously a combination of two elements that can usually appear independently as well. Net as lon as there is no definite evidence to the contrary, writers may choose to assume that even in the perfect tense, the initial consonants of verbs were normally re&formed by analo y with the unau mented shape of the verb. .e donLt have to ma"e our 5eo&Quenya more complicated than we definitely "now Tol"ien ima ined his Quenya to be.

A somewhat similar problem has to do with the %superlative or intensive prefix% an& that may be prefixed to adMectives ((etters#H?3). As pointed out in (esson @ive, the n of this prefix must be assimilated to the followin consonant when it is l&, r&, s&, or %&, e. . li"e this# an J lauca %warm% : allauca %warmest% an J rin?a %cold% : arrin?a %coldest% an J sar!a %hard% : assar!a %hardest% an J %oina %dear% : a%%oina %dearest% This system I ori inally had to deduce from eneral principles, since for years ancali%a %bri htest% was the sole available example of the prefix an&. These assimilations were finally confirmed in material published in CT8I#I, 0>. A parallel example involved the prefix lin& %many% (itself listed in the entry LI& in +tym)# In an adMective mentioned in 7,#HH0, namely lillassa %many&leafed, havin many leaves%, the &n of lin& is seen to be assimilated to l before another l (compare lass %leaf%). $owever, one example of lin& occurrin in the Etymologies is of particular interest# @rom the noun norno %oa"% is derived the adMective lin!orna %havin many oa"&trees%. .hy lin!orna and not **linnorna, if norno is the basic word here; It is because norno %oa"% is derived from a stem /9.9N (under which entry these words are listed in +tym). Norno is one of the exceptional cases where the initial d& of primitive +lvish produces n& instead of l& in Quenya. $owever, followin the prefix lin& the ori inal ! is still preserved in Quenya# perhaps the adMective lin!orna oes all the way bac" to primitive lin0 doron2y or whatever. @or whereas initial d& was eventually altered to l& or n& in Quenya, the combination nd occurrin between vowels survived unchan ed. In the ori inal version of this appendix, I as"ed# The Buestion before us is this# would the same principle apply to the superlative prefix an&; Ta"e the adMective norna %stiff, tou h%, which Tol"ien derived from a stem /9.9 (./#840&48). If we supply the prefix an&, should %tou hest% be an!orna rather than annorna; Aimilarly with adMectives in l& ori inatin from d&# Ahould a word li"e lu%na %ominous%, derived from a stem /1% in +tym, have the superlative form an!u%na because of this derivation; If so, one must in each case "now whether the initial l& or n& of an adMective is ori inal or chan ed from d& before one can construct the proper superlative form. These theories were finally confirmed by Tol"ien material published in CT8I#0>. Tol"ien does note how an& is normally assimilated to al& before another l, but in the case of l derived from ori inal !, we et an!& instead. Ao iven the fact that lu%na %ominous% descends from a primitive form that had d&

instead of l& as the first consonant, %most ominous% would indeed be an!u%na, not **allu%na. In CT8I, Tol"ien does not comment upon n& derived from !& (as in the case of norna %tou h%), not surprisin since this is a rare development. But the eneral principle does point to an!orna rather than **annorna as the word for %tou hest%. Aimilar complications could occur in the case of initial %& or n& where they are simplified from ori inal m#& or nd&Q3g& (cf. our discussion of au mented perfects above). The adMective %ar a %fey, fated% is evidently meant to come from primitive m#arat (the stem M%,.," is listed in +tym), and it may well be that an0m#arat would come out as a%,ar a in Quenya. The adMective nulla %obscure% is explicitly said to come from earlier ndulla (entry N/1L in +tym), and then %most obscure% should perhaps be an!ulla (for an0 ndulla) rather than annulla. .ords that ori inally had initial 3g& still showed =& (as described above) in the early %historical% or %recorded% forms of Quenya, thou h it had mer ed with n& in Third A e pronunciation. ,onsider adMectives li"e n'la %learned% or n.alca %cruel% (stems 459L, 45:,L)# older ='la and =.alca. Applyin the prefix an& would li"ely produce an?'la and an?.alca, whether we start from primitive an03g2l, an03gwal or later an&='la, an& =.alca (for n= would in any case assimilate to ==, and this combination also produces Quenya n? or technically =?). The material published in CT8I does not discuss what happens when an& is prefixed to words in n& and %& derived from primitive nd&, 3g&, m#&. Accordin to CT8I#0>, the otherwise lost initial g& of Quenya words is preserved followin the superlative prefix. Thus words li"e aira %ruddy% (stem 5,6) and .enya % reen% (stem 5:EN) may become angaira, ang.enya if the prefix is added to the word.
Actually CT8I#0> rather points to in?aira etc., since it is there said that the prefix has the vowel i before c and (ori inal) ? 6 but since Tol"ien later let %most bri ht% be ancali%a rather than **incali%a, this idea must have been abandoned. In CT8I#I Tol"ien refers to how the prefix was enerali!ed in forms with the vowel a.

AdMectives in #& may also be problematic. Aometimes #& comes from primitive #&, sometimes from w&. In the oldest %recorded% form of Quenya, w& still remained unchan ed, so the ori inal distinction #& vs. w& was preserved as #& vs. .&. (The words in ori inal .& are often so spelt by Tol"ien, e. . .en! besides #en! for %maiden%.) It seems that even after initial .& became #&, the combination n. remained unchan ed- for instance, there is no hint that a word li"e an.a %real, actual, true% ever became **an#a. Therefore, I ori inally noted that %one could ar ue that an adMective li"e #3ra %private% would become an.3ra if the prefix an& is applied, since in L=ld QuenyaL this adMective appeared as .3ra ('7#089)%. 7aterial from CT8I#0> now confirms that an& J (ori inal;) . produces an.&, thou h the shift of initial .& to #& is not there discussed. Therefore, there may be an openin for analo ical an#& at a later sta e (see below).

An adMective with #& derived from ori inal #& could behave Buite differently. Varna %secure% comes from an ori inal #arn (stem %,.), and an0 #arn could well come out as a%,arna in Quenya, with assimilation n# K %,. Ao to correctly apply the prefix an&, maybe one has to "now whether the initial #& of a Quenya adMective comes from #& or w&; At least there can be no doubt that adMectives in .& would originally show an.& if the prefix an& is applied- this is now confirmed by CT8I#0>. The real Buestion is whether the prefix an& can simply be added to the %contemporary% forms in #&, without havin to consider the whole underlyin historical scenario. Aince .3ra %private% eventually became #3ra, could %most private% simply be an#3ra, irrespective of the phonolo ical history that would demand an.3ra; 'erhaps. .e have one parallel example involvin the prefix en& %re&, a ain%. In the attested words en#inyan a %healed, renewed% (72#89I) and Ara ornLs title En#inya ar %2enewer%, it is simply prefixed to a word incorporatin the contemporary form of #inya %new%. If #inya is meant to come from #iny, then Ara orn %should% perhaps have called himself **E%,inya ar. Alternatively, if #inya is to be derived from winy (and Tol"ien did ima ine a root :IN, CT8>#H>), then the historically Mustified form %should% have been **En.inya ar. The attested forms En#inya ar, en#inyan a su est that one does not always have to consider the whole evolution supposed to underlie the words as they appear in Third A e Quenya. There is no reason to assume that the prefixes en& and an& would behave differently in this respect. Ao the word for %newest% can evidently be simply (an& J #inya :) an#inya. +ven if it was an.inya at the older sta es, the prefixed form could also be altered when .inya became #inya (even thou h n. remained in unitary words li"e an.a %true%). This principle may be relevant for many of the potential problems here discussed. .e may also consider the primitive aspirated consonants h, th, ph, pronounced more or less as in as in #ackhand( outhouse, scrap-heap (to borrow my own examples from (esson =ne). In Quenya, initial h& first became ch as in <erman ach- later it was wea"ened to a sound li"e +n lish 0&. =ri inal initial ph& became Quenya $&. "h& first came to be pronounced somewhat li"e +n lish th (as in thin )- later this sound mer ed with s&. Thus primitive forms li"e hithw, phirin&, thaus yielded the Quenya adMectives 0is.a % rey%, $irin %dead% and saura %foul% (related to the name Sauron). $owever, followin at least some consonants, the primitive aspirates h, ph, th lost their aspiration (the h element) and became unaspirated , p, t instead. These sounds normally survived unchan ed into Quenya (8 here spelt c). This loss of aspiration did occur followin n, as when Tol"ien derived the Quenya verb %anca& %trade% from a root M%,K- %exchan e%# +vidently this m#a h& early underwent nasal infixion, and n h is seen to have become n8 (nc) in %anca&. If the same thin were to happen followin the superlative prefix an&, then an0 hithw, an0phirin& and an0thaus would come out as Quenya ancis.a, (an2irin K) a%2irin and

an aura# rather unexpected forms compared to 0is.a, $irin, saura without the prefix. This system would also lead to confusion with other words# are the Buite distinct adMectives aura %mi hty% and saura %foul% to share the superlative form an aura;O It may be noted that in the entry $-I. in +tym, from which stem the adMective $irin %dead% is derived, Tol"ien also mentioned a word for %immortal%# il$irin. $ere, $irin is supplied with the ne ation il&. Interestin ly, Tol"ien noted that the form il$irin occurred %for *ilpirin%. In other words, followin l the ori inal ph& of the stem $-I. would normally have become p, so that the Quenya word %should% have been il2irin, but Tol"ien asteris"ed this form to indicate that it was not in use. +vidently it was re&formed as il$irin by analo y with the independent word $irin. Then it would probably also be permissible to let adMectives retain their normal initial consonant even when the superlative prefix an& comes before it. If we start considerin how the underlyin phonolo ical evolution could affect inflected forms in Quenya, yet another possible complication has to do with nouns in &il that is descended from primitive +lvish &la. Tol"ienLs idea was that a primitive word li"e ma la %sword% first became ma l, since final short &a disappeared Buite early. This ma l was pronounced as two syllables, ma 0l with a syllabic ( at the end (Must li"e +n lish little is pronounced lit0l). +ventually, a new vowel i developed before this syllabic (, so that the word came to end in &il. In this way, an ori inal ma la evolved into Quenya %acil. =ther words of this "ind include ecil %pen% and 0ecil %waif, outcast% (primitive te la, he la). But what happens if such words are to receive case endin s; If, for instance, we are to add the instrumental endin &nen to %acil to express %with a sword%, should we start from the primitive form ma la0nen; This combination would develop differently. In ma la0nen the final &a of ma la is not final at all, and then it would presumably not be lost. =ne could ar ue that in Quenya, the historically Mustified instrumental form %with a sword% should be somethin li"e %alcanen 6 since in Quenya the cluster l occurrin between vowels turned into l : lc. (@or instance, alcar % lory% comes from primitive a lar0- compare Aindarin aglar.) Aimilarly, %with a pen% ( ecil) could be elcanen, since ecil descends from primitive te la. .hen constructin case forms, is it permissible to ta"e the simpler route of simply startin from %acil, ecil etc. and inflectin them li"e other nouns; (Then the instrumental form of %acil would be either %acilenen with a connectin vowel before &nen, or perhaps %acilanen with the otherwise lost final vowel of the old endin &la preserved, or even %acil!en for macilnen with the normal development ln K l!.) It may be noted that after derivin the Quenya noun 0ecil %outcast, waif% from primitive he la, Tol"ien also mentioned masculine and feminine forms of it# masc. 0ecilo, fem. 0ecil (./#0>I). These words seem to be derived simply by addin the masculine and feminine endin s

&o, & to the word 0ecil as it exists in the %contemporary% form of Quenya# The masculine endin &o probably descends from primitive &2, and primitive he la02 mi ht come out as **0elco in Quenya 6 but this form does not occur. If derivational endin s can be added to the noun 0ecil with no further complications, then it is perhaps also permissible to attach case endin s to these nouns in &il with no stran e thin s happenin to the noun&stem itself. A ain, it may not be necessary to ta"e into account the entire historical development that is supposed to have produced the %contemporary% (synchronic, as lin uists mi ht say) form of the noun. As I have already indicated, there is currently no definite answer available to many of the Buestions raised above. .e have merely s"etched some of the potential implications of Tol"ienLs eneral phonolo ical system, which is inseparable from his vision of how Quenya had evolved from the earlier forms of +lvish. As we have seen, there are hints that analo ical levelin would sometimes be permissible, but also examples that point in the opposite direction. =nly future publications can throw more li ht on these problems, assumin that Tol"ien discusses such matters in his manuscripts. There will certainly be some clues, iven Tol"ienLs deep interest in the historical evolution of his lan ua es (which to him often seems to be more important than their %modern% or synchronic manifestation). AT+7 CA2IATI=5 In (esson Three of this course, one important fact about Quenya was introduced# ,ompared to the endin less form (the simple; form) of a word, the word will in some cases subtly change when endin s are added to it 6 as when the plural form of alan %floor% is ala%i rather than ** alani. Thus the simplex alan has the stem0form ala%&. Typically, the stem&form reflects how the word appeared at older sta es of the lan ua e (for instance, the simplex talam was altered to alan because Quenya at one sta e no lon er tolerated final &m and chan ed it to the closest permissible sound, namely &n). I wrote, reassurin ly# %The student should not despair, thin"in that all sorts of stran e thin s typically happen whenever you add an endin to a Quenya word, so that there is a reat potential for ma"in embarrassin mista"es (or at least very much extra stuff to memori!e). 7ost Quenya words seem to be Buite well&behaved, with no distinct LstemL form to remember- you Must add the endin and thatLs it.% .e will now attempt a fuller survey of the exceptional words that have special stem&forms. Tol"ien himself referred to these stem&forms in various ways. Aometimes he listed them as we do here, with a hyphen where the endin is to follow, e. . n3n, nen& %water% (+tym, entry NEN). This indicates that, say, the plural %waters% is to be neni rather than **n3ni. At other times, Tol"ien himself mentions a complete inflected form, most often the plural or the enitive (which in the

Etymologies still has the endin &(e)n rather than &o- in (ot2&style Quenya, these forms as such are presumably still valid but must be ta"en as dative forms instead). The fact that alan %floor% has the stem ala%& Tol"ien indicates by citin the plural form ala%i (see +tym, entry ",L, ",L,M). In the same entry he also mentions the word )l %foot%. Before endin s this word appears as al& with a short vowel, and to indicate this, Tol"ien cited the % .s .% ( enitive sin ular, later evidently dative sin ular) form alen. Aummari!in various roups of stem&forms, a picture somethin li"e the followin emer es. Quite a few nouns that have simplex forms in &n, &r or &l have stem&forms that add a &!& to this consonant# R an!on % reat ate% K an!on!& R car %buildin , house%, also %deed% K car!& R $al%arin %nymph, sea&spirit% K $al%arin!& R $ion (%haw"%; & "ol ien+s gloss was illegi#le) $ion!& 7#ut also simply fion08 R 0en %eye% K 0en!& R 0.an %spon e% K 0.an!& R Laurelin (name of the <olden Tree of Calinor) K Laurelin!& 7#ut also Laureling-<
see #elow8

R L'rien (place&name) K L'rien!& 7locative Lriendess in .5E9=>>8 R %eren %feast% K %eren!& R nel il %trian le% K nel il!& R '%an %vowel% K o%an!& 7e?g? pl? omandi< misprint or misreading @amandi@ in the entry =7 in
Etym as pu#lished in L.A nonetheless< it may #e correct that the initial long vowel #ecomes short8

R 2ilin %arrow% K 2ilin!& R sar %(small) stone% K sar!& R Solonel %Teler&+lf% K Solonel!& R Tani9ue il (name of mountain) K Tani9ue il!& 7#ut in the a#lative< the contracted form
Taniquetillo is perhaps a #etter choice than Taniquetildello8

R .en % irl, maiden% K .en!& R .il.arin %butterfly% K .il.arin!& 7in this and the previous word< read perhaps v0 for older
initial w0 in late E;ilic 'uenya8

In Namri we find the word oro%ar!i, translated %hi h halls%. If %ar!i is actually the plural form of %ar %home% (cf. car, sar pl. car!i, sar!i), the word %ar also belon s to this cate ory. But %ar!i may also be the plural form of a distinct word %ar! %hall%, not otherwise attested. )oes the endin &on that often occurs in masculine names (e. . Ancali%on, Sauron) become &on!& before an endin ; Tol"ien informs us that Sauron comes from older T0auron! ((etters#019). $ence enitive Sauron!o, dative Sauron!en; $owever, the patronymic endin s &ion %son% and &iel %dau hter% are evidently unchan ed (i.e., do not become **&ion!& and **&iel!&), thou h

they correspond to the independent words yon!o, yel!# In '7#43H, 43> cf. 884, we find Isil!urioni rather than **Isil!urion!i for %the heirs Flit. sonsG of Isildur%, and li"ewise An)rioni rather than **An)rion!i for %the heirs of AnSrion%. Therefore, the patronymic endin s &ion %&son% and (presumably li"ewise) &iel %dau hter% are unchan ed before rammatical suffixes. These endin s probably represent the simplest forms of the roots 69N, 6EL, while the independent forms yon!o, yel! include a %stren thenin % of the medial consonant (n becomin n! and l becomin l!).
5ote# The endin &riel in the name Al ariel %<aladriel% has nothin to do with &iel %dau hter%- it is a contracted form of riell % arlanded maiden%, so this name becomes Al ariell& rather than remainin AAAl ariel& before an endin ( enitive Al ariello %<aladrielLs% in 2<+=#>>)

Another freBuent endin in names, &!il or &nil %friend, lover%, li"ewise does not become **&nil!& or **&!il!& before an endin , even thou h it corresponds to the independent words nil!o, nil! %friend% (masculine and feminine, respectively). This is evident from the form Nen!ili %.ater&lovers% in ./#844 6 not **Nen!il!i. =nce a ain, the endin must be assumed to represent the simplest form of the ori inal root NIL, N/IL. A few stems add a (or historically spea"in , a final & has been lost in the simplex forms)# R oron %mountain% K oron & R u%,ar %faith, doom% K u%,ar & (and similarly in the name Tura%,ar, Tura%,ar & containin
a%,ar( )& as another word for %doom%, not related to A%,ar %world%- the instrumental form of a%,ar %doom% is attested as a%,ar anen)

R *an!os (usual name of the Cala 5Smo, properly the place where he dwells) K *an!os & Aince the final element of coi%as %life&bread, lembas% is a reduced form of %as a %bread% (+tym, entry M%,!&), coi%as could become coi%as & before endin s. ,ompare *an!os, *an!os &- the final element of this name is a shortened form of os o %castle, fortress- city%. Net Tol"ien in his later writin s also experiments with %assa (not %as a) as the word for %bread%, and then coi%as would perhaps have the stem coi%ass& instead. Aome stems are contracted. This simply reflects the re ular Quenya %syncope% (the process whereby the second of two identical short vowels normally disappears in a word of more than two syllables, as when primitive galad %tree% produced Quenya al!a- notice how the second a of the primitive word has dropped out). $ere, the stem&form does not reflect the older shape of the word- it has been shortened. This list hopefully covers most of the contracted stems mentioned in Tol"ienLs published writin s# R coron % lobe, ball% K corn& R $eren %beech&tree% K $ern&

R 0aran %chieftain% K 0arn& R 0uan %hound% K 01n& R la%an %animal% K la%n& 7#ut also uncontracted laman08 R nele %tooth% K nelc& 7with #oth contraction and variation tBc0A see #elow8 R seler %sister% K sell& 7for older selr-< or historically actually sels0 since the stem is T$+(+A8 R soron %ea le% K sorn& R oron %brother% K orn&
5=T+# Aome would re ard the nouns seler %sister% and oron %brother% as obsolete, since in a much later source Tol"ien cited wholly different Quenya words# n3sa %sister% and 0)no, 0anno %brother% (CT8?#48).

In poetry, the noun elen %star% mi ht sometimes appear as a contracted form# el!& (e. . pl. el!i, representin older elni Fstill so in the Telerin dialectG- the Quenya development ln K l! is re ular). $owever, by analo y with the simplex form, this noun normally preserved the full stem elen& even where it occurs with endin s (hence pl. eleni in Namri). Aee ./#0>H.
5=T+# The possessive endin &#a would probably appear as &.a when added to nouns endin in a consonant. It is probable that contracted stems would not be used when this case endin is added 6 e. ., the possessive form of oron %brother% would most li"ely be oron.a, hardly ; orne#a with a contracted stem J connectin vowel &e&. Addin &.a after the final consonant of a word creates a consonant cluster (li"e the n. of oron.a in our example), and then the syncope of the precedin vowel cannot ta"e place ( oron.a cannot possibly become ** orn.a). The same principle would li"ely apply to short locative forms in &s (instead of the full endin &ss, which would always reBuire a connectin vowel before it if a vowel is not already present). As we explained in (esson @ifteen, the shorter endin &s may apparently be added to nouns in & and &s, and followin nouns in &n and &l, the locative endin may appear as &! (chan ed from &<, in turn from ori inal &s)- thus Tol"ien used %enel! and ce%en! as the locative forms of %enel %heaven% and ce%en %earth% (CT80# 4>, 4?). Thou h coron %ball% is normally contracted to corn& before endin s, the locative %in a ball% should perhaps be coron! 6 thou h corness may be a valid alternative. But the instrumental should probably be coronnen, hardly ; cornenen.

Aome stems endin in &s are "nown to double it before endin s# R e9ues %sayin % K e9uess& R lis %honey% K liss& R n"s %woman% K niss& R Tul8as (the name of a Cala) K Tul8ass& 5otice that the lon " of n"s becomes short before a consonant cluster in niss&. (The simplex form n"s occurs in 72#H40, perhaps renderin obsolete the short& vowel form nis mentioned in the Etymologies, entries N/I!, NCD< NI!.) 7iscellaneous stem&variation# R cai%asan %bedchamber% K cai%asa%,& 70san in the first word #eing a reduced form of
samb @cham#er@? 9ther compounds ending in 0san @cham#er@ would #ehave in the same way8

R $ili %small bird% K $ilic& 7the stem originally ended in c< #ut as the language developed< word0final c
was no longer possi#le and was changed to t E #ut in non0final position< @shielded@ #y an ending< it remained c? Fompare nelet< nelc0 #elow?8

R 0ala ir %"in fisher% K 0ala irn& 7with an e;tra 0n #ecause the second element of the name is
derived from tirno @watcher@A the 'uenya name of the #ird means @fish0watcher@8

R Laurelin %(.% K Laurelin?& 7or Laurelind0A see #elow8 R %iril %shinin Mewel% K %irill& R nele %tooth% K nelc& 7with #oth contraction and variation tBc0A the #ase yielding words for @tooth@ is
given as 5T(&+U0 in Etym8

R noa, n' %conception% K n'.& 7or may#e nw0 is the stem0form of n onlyA "ol R 2el as %pivot% K 2el a@& R 9uele %corpse% K 9uele & 7"ol

ien+s wording is not clear? -e cited the plural of n as nwi< whereas noa would presuma#ly have the plural form noar?8 ien listed an archaic form kwelett0 and the 'uenya plural queletsiA this plural reflects the idea that t #efore an unstressed i may #ecome s? Ff? another form from the +tymolo ies= maisi as the plural form of the ad&ective mait @handy@ E #ut there are some indications that "ol ien later dropped this idea< so perhaps the plural of quelet could also simply #e queletti?8

R 9uesse %pillow% K probably 9uessec& 7since the co

nate or corresponding @Noldorin@B!indarin word< pesseg< indicates that the primitive form is "wesse"-A cf? filit with stem filic0 from the stem $-ILIK< with @N@B! cognate fileg8

R r) %lion% K r)#& R Sil%aril %Ailmaril% K Sil%arill& R alan %floor% K ala%& 7a case similar to the change of c to t mentioned a#oveA originally< the stem
always ended in m8

R ' %wool% K probably '.& 7from T=.A cf? n< nw- from 5=.= a#ove8 R ol %island% K oll& 7pl? @tolle@< according to Etym< entry T=(H? !ince there are no other e;amples of
any - plurals< and #ecause L"D=GH has the more regular form tolli< it would seem that @tolle@ is pro#a#ly a mere misreading or misprint?8

R ya %nec"% K ya0 & The name of the <olden Tree of Calinor, Laurelin, was usually interpreted %Aon (lin!) of <old% and had the stem Laurelin!& before an endin . But the name was also interpreted %$an in &<old% (cf. lin?a& %han %) and conseBuently became Laurelin?& when an endin was added to it (+tym, entry LINI). .riters may pic" the form they prefer. Aince a%il %mother% seems to be shortened from a lon er form a%ill (CT88#?), it is probable that a%il should have the stem a%ill&, e. . enitive a%illo %motherLs%. The forms in & with stem&forms in &c& may reBuire special attention. .e have mentioned $ili %small bird% ($ilic&), nele %tooth% (nelc&) and 9uesse %pillow% (probably 9uessec&). If we add the endin &.a for the possessive case, the combination c0w would eBual 9u, so we would have (filic0wa :) $ili9ua %of a (small) bird%, nele9ua %of a tooth% (probably no contraction here, thou h nel9ua as such would be a possible Quenya word) and 9uesse9ua %of a pillow%. If we add the short locative endin &s, the resultin combination c0s would have to be spelt @ accordin to the ortho raphy here adopted# $ili@ %in a bird%, nele@ %in a tooth% (definitely no contraction here, since **nel; is impossible), and 9uesse@ %in (Qon) a pillow%. But in all li"elihood one could also slip in a

connectin vowel &e& and use the full locative endin &ss# $ilicess, nelcess, 9uessecess. It may also be mentioned that nouns with stem&forms in &%& (for &n in the simplex forms) would probably still show &n& before the possessive endin &.a# =lder mw came out as n. in Quenya (see CT84#I, Tol"ien derivin the noun san. %thou ht% from older sam0wJ). Therefore, the possessive form of alan, ala%& %floor% would presumably still be alan.a, not ** ala%.a. .hat about the locative; Dnless one says ala%ess with a connectin vowel before the lon endin &ss, talam0se with the shorter endin would presumably evolve into talamze and then alan!, a ain with the same chan e of % to n that also occurs in the simplex form, thou h for a different reason# It seems that older md becomes n! in Quenya, as when Tol"ien in +tym derived 2ilin!i %arrows% from the root $KLIM. This plural form must reflect older pilim0d0L, the roup md becomin n!. M9:EL M,.I,"I9N! Aome nouns shorten a lon vowel before an endin (or, more accurately historically spea"in , lengthen it when no endin is present)# R n3n %water% K nen& R n3r %man% K ner& R 9u3n %person% K 9uen& Accordin to the entry 6EN of +tym, the lon vowel of y3n %lon year% becomes short before an endin (yen&), but in Namri the plural form is y3ni with the lon vowel intact (so also in CT88#00, in Tol"ienLs fra mentary Quenya translation of the 5loria in E;celsis /eo- here we also have the enitive plural y3nion). It would seem that Tol"ien altered the ultimate etymolo y of the word, so that it had a lon vowel already at the primitive sta es of +lvish speech. If we want to ar ue from an %internal% perspective, we could also theori!e that in late +xilic Quenya, the older system was brea"in down so that the lon vowel was introduced %everywhere% by analo y with the simplex form. If so, <aladriel would perhaps also be liable to use plurals li"e ;n3ni, ;n3ri, ;9u3ni for %waters, men, persons%, thou ht this should have been neni, neri, 9ueni accordin to the older, classical system (the forms neri and 9ueni are directly attested). There are also a some words where the vowel in the final syllable of the simplex form becomes long before endin s# R Eru0in %,hild of +ru% K Eru0"n& (as in pl. Eru0"ni- cf. the lon vowel of the independent word 0"na %child%) R Vala ar %Cala&"in % K Vala )r& (pl. Vala )ri- cf. the lon vowel of )r %lord, "in %)

R Casar %)warf% K Cas)r& (pl. Cas)ri, but also simply Casari, ./#011, 013, 89H) The two first examples demonstrate how the lon vowel of a word is normally shortened when the word (or a reduced form of it, li"e &0in vs. 0"na) occurs at the end of a compound. Before endin s, the lon vowel is preserved. As for Casar, it is supposed to be borrowin from the )warvish (Uhu!dul) term Khazd, itself a plural form %)warves%. (In actual conversation with )warves, Casar is supposed to be a more polite and politically correct term than Nauco, which is derived from the adMective nauca %stunted%.) It is evidently the lon of the Uhu!dul term that is reflected in the Quenya plural Cas)ri. A ana )ri %@athers of 7en% ('7#0H8) is the plural of A ana ar %7an& father%, this sin ular form bein attested as the personal name of one of the "in s of <ondor (Appendix A of (ot2). 5otice the len thenin of the ) in the plural form (as well as in the enitive plural A ana )rion, 72#0?0 6 in which source this is the title of a collection of le ends and translated %(e endarium of the @athers of 7en%, but the word %(e endarium% is understood in Quenya). )o these examples indicate that the plural form of a ar %father% is re ularly a )ri, Must li"e the plural of Casar is Cas)ri, so that a ar would have the stem a )r&; 7aybe not, for in the Etymologies, the plural form of a ar is simply a ari (entry ,",&). .hile Tol"ien may have chan ed his mind about this later, it is possible that a lon compound li"e A ana ar %7an&father% does not behave in the same way as the simplex a ar %father%. A plural form ;A ana ari would have to be accented on &nat&, which would sound rather aw"ward- perhaps this is why the vowel in the followin syllable is len thened so as to receive the stress# A ana )ri. It may be noted that in some early sources, written before Tol"ien chan ed the enitive endin from &(e)n to &o, the enitive of Il1#a ar %All& father% was Il1#a )ren ((2#8?, ?H). $owever, in the entry %,hildren of IlPvatar% in the !ilmarillion index, the revised enitive of the same name appears as Il1#a aro with no len thenin of the a in the second&to&last syllable. =ne must wonder if this should read ;Il1#a )ro (if only because Il1#a aro would presumably be stressed on &vat&, a rather weird accentuation). Anar %sun% was ori inally ,nr with a lon vowel in the second&to&last syllable (+tym entry ,NN., also A)#09>). Accordin to what Tol"ien wrote in the 'lot! (etter, the shortenin of lon vowels in the final syllable of polysyllabic words is a feature of +xilic Quenya. It may be that it is this ori inally lon vowel that is preserved in the name An)rion (son of +lendil and brother of Isildur)- the name seems to mean %Aun&son%. .ould the enitive plural %of suns% li"ewise be an)rion, in turn ar uin the existence of a plural form an)ri, a enitive form an)ro, etc; .e cannot be certain. In a number of three&syllabic words, a lon vowel in the middle syllable is actually seen to be shortened (so that the stress moves to the initial syllable of the word). @or instance, Quenya naraca %harsh% Tol"ien referred to older nar a (+tym, entry

N,.NK). If nar a can produce Quenya naraca rather than **nar)ca, perhaps a plural form li"e ;an)ri %suns% would li"ewise tend to become anari. The fact that a lon vowel persists in Cas)ri %)warves% is easily explained# this is presented as a late loan from Uhu!dul Khazd, not an inherited word. Indeed even Cas)ri could be replaced by Casari (and presumably li"ewise in other forms involvin endin s, e. . Cas)ro or Casaro as the enitive sin ular %)warfLs%). Aee ./#011. Aometimes the sound of a final vowel is different before endin s (one vowel chan in into another, not Must the same vowel bein len thened or shortened as in the cases discussed above). .ords in final & and &o sometimes display stem variation when endin s are added# .here these vowels are derived from final short &i and &u in primitive +lvish, we still see &i& and &u& where these vowels are followed by an endin . This occurs all the time in the aorist forms of primary verbs# o2 %covers% vs. o2in %I cover% (the form o2 descendin from older topi). In the case of ad&ectives in &, nearly always representin earlier &i, the ori inal Buality of the vowel would also be preserved in many cases. It often occurs when adMectives are compounded, as in %or %dar", blac"% vs. *ori9uen!i %)ar" +lves% (see the !ilmarillion for the latter term). There are some exceptions- nin9u %white% comes from primitive nin wi and therefore ou ht to have the stem&form nin9ui&, yet we see nin9ue& in the name Nin9uel' %.hite @lower% (see the !ilmarillion Appendix, entry loth&). In the case of carn %red%, Tol"ien in the first edition of (ot2 simply used carne& in the compound Carne%"ri %2ed& Meweled% (the rowan&tree in the +nt Quic"beamLs son - see "he "wo "owers, chapter four in Boo" Tree- compare the entry caran in the !ilmarillion appendix). Carne%"ri is also the readin found in (etters#HH8. Net carn is derived from primitive arani and therefore ou ht to have the stem&form carni&. Apparently reali!in this, Tol"ien in the revised second edition of (ot2 emended Carne%"ri to Carni%"ri. The form Nin9uel' remains seemin ly abnormal, but nin9ui& as the stem&form of nin9u %white% is attested in the word nin9uil!a %whiter%. This form was published in "yali "yelelliOva V4> p. H8, where (isa Atar presented evidence for one form of Quenya (or %Qenya%) that had the comparative endin &l!a. These adMectives in & are "nown to have, or may safely be inferred to have, stem&forms in &i&# R carn (carni&) %red% (primitive arani, +tym entry K,.,N) R $or ($or i&) %northern% (;) (primitive phoroti, whence %5oldorin%QAindarin forod accordin to the entry $-9. in +tym- however, the more usual Quenya word for %northern% seems to be formenya)

R l1n (l1ni&) %blue% (primitive lugni, +tym entry L15I 6 but in Namri the plural form of the adMective %blue% appears as luini, perhaps s . luin, but Buite possibly this should also have the stem luini&) R %ai (%ai i&, or %aisi&) %handy, s"illed% (primitive maPiti, +tym entry M,P) R %or (%ori&) %blac"% (primitive mori, +tym entry M9.) R nin! (nin!i&) %fra ile, thin% (the entry&head NIN0/I& in +tym would seem to represent a primitive word nindi) R nin9u (nin9ui&) %white% (primitive nin wi, +tym entry NIK0:0) R n" (n" i&, or n"si&) %moist, dewy% (primitive neiti, +tym entry NEI) R rin? (rin?i&) %cold% (the entry&head .IN5I in +tym apparently represents a primitive word, not Must a root 6 but later, Tol"ien seems to have altered the Quenya word for %cold% to rin?a) R sin! (sin!i&) % rey% (primitive thindi, '7#018 and +tym entry "-IN) R #arn (#arni&) %brown, swart, dar" brown% (Tol"ien explicitly mentions the stem&form in +tym, entry %,.NN, so the primitive form would be #aranicompare carn above) R #in! (#in!i&) %pale blue or rey% (primitive form windi iven in +tym, entry :INQ:IN/) It seems that virtually all Quenya adMectives in & represent primitive forms in &i and would belon on this list ( eren %slender% from primitive terJnJ is the only exception that sprin s to mind). 7any of the adMectives in &i denote colours, as is evident from the list above. ,ertain nouns in & may also have stem&forms in &i&, that would turn up before endin s and when such nouns appear as the first element of a compound. This would be a partial list of such nouns (not includin Tol"ienLs early %Qenya% material)# R e0 (e0 i&) %spear% (since in +tym, Tol"ien decided to derive this word from EK"I rather than EK"E as he first wrote) R $in ($ini&) %a sin le hair% (primitive phini, '7#0>H- cf. Finic)no as @in onLs Quenya name in some drafts, '7#0>4 no. 0I, thou h Tol"ien later altered this to Fin!ec)no usin another word for %hair%) R 0"s %mist, fo % (primitive hLthi, +tym entry K-I!, K-I"-- cf. also the compound H"sil'%, a place&name mentioned in the !ilmarillion) R lin. (lin?.i&) %fish% (primitive li3wi, +tym entry LI:- cf. the compound lin?.il'c %fish&dra on, sea&serpent%, +tym entry L9K) R l"r (l"ri&) %son % (instrumental l"rinen in Namri) R l'% (l'%i&) %ni ht% (primitive d2mi&, +tym entry /9M9- A)#84I explicitly confirms the stem&form) R nen?. (nen?.i&) %nose% (the entry&head NE40:I& in +tym seems to represent an entire primitive word, not Must a %stem% or root)

R noir (noiri&) %tomb% (compounded in Noirinan %Tomb&valley%, %Calley of Tombs%, DT#4>> 6 unless this compound is meant to contain the plural form noiri %tombs%) R 2or (2ori&) %flour, meal% (primitive pori, +tym entry $9.) R s1r (s1ri&) %wind% (instrumental s1rinen in Namri) R rinc (rinci&) %flourish, Buic" sha"e% (primitive rin i, +tym entry .IK7-8) It should be noted thou h, that ad&ectival formations derived from these words by addin &a to the final vowel do not cause & to revert to &i&. This is evident from the example nen?.a %nasal%, an adMective derived from nen?. %nose% (+tym, entry NE40:I&). 'erhaps the %expected% form **nen?.ia was altered by analo y with the numerous other adMectives in &a. In the Etymologies, the Quenya word for %day% is ar derived from ari (,.D), so the stem&form would be ari&, but Tol"ien later chan ed the word for %day% to aur. .hether this has the stem&form ;auri& is uncertain and possibly doubtful. The list of nouns above only comprises the examples where Tol"ien explicitly mentioned the primitive form in &i, or where the i&stem can be directly observed in some inflected or compounded form. =f course, there are many words in & for which no primitive form is cited, and then we cannot always be certain whether Quenya & derives from short &i (in which case it belon s on the list above) or from earlier lon &J (in which case Quenya & does not chan e its Buality when not final). It may be assumed, thou h, that the noun allun %sole of foot% is an I&stem ( alluni&), since it derives from older talrunya (+tym, entry .1N). The idea is probably that talrunya produced ,ommon +ldarin talruny after the loss of the short final &a, and that the consonant y then turned into a vowel &i (talruni, the form directly underlyin Quenya allun). ,ompare the apparent evolution of the D&stem an?o, an?u& from angwa via angw and angusee below. As a parallel to the I&stems Must listed we have the 10stems, showin &o as the final vowel of the simplex form, but &u& before endin s. .here the simplex form of such nouns ends in &co or &?o, the nominative plural ends in &9ui or &?.i, respectively (instead of &cor, &?or as would normally be the case). ,ompare ./#039, where Tol"ien states that urco %bo ey- =rc% must descend from ur u or uru u because it has the plural form ur9ui. The implication is that if urco had descended from ur7u8 2 instead, its plural form would have been **urcor instead. In the case of the noun ulco %support, prop% derived from tul u (+tym, entry "1L1K), Tol"ien did not ma"e any note about the plural form, but iven the derivation it should probably be ul9ui rather than ; ulcor (or for that matter ; ulcur).

5=T+# Telco %le % is a special case- it has the plural form el9ui, but this form Tol"ien stated to be %analo ical% (+tym, entry "QLEK). It seems, then, that elco is not really a D&stem- the plural el9ui is merely formed by analo y with the plural of certain true D&stems. Ara ornLs Quenya translation of %Atrider%, Telcon ar, seems to somehow contain elco %le % (it may be that a verb elcon a& %use the le % : %stride% is implied here), and the fact that we donLt see **Telcun ar would then confirm that elco %le % is not a D&stem despite its plural form.

As noted above, Must li"e nouns in &co derived from earlier & u have plural forms in &9ui (: &c.i), nouns in &?o derived from earlier &gu have plural forms in &?.i. The noun an?o %sna"e% has the stem an?u&, as is evident from the compound an?ul'c %dra on% in +tym, entry L9K (literally somethin li"e %sna"e& serpent%, compound of an?o and l'c), and the entry ,N5:, confirms that an?o has the plural form an?.i. (In this case the &u of the stem seems to have developed from an ori inal &w, which became final after the loss of the ori inal short final vowel &a in ,ommon +ldarin# angwa K angw, later angu, in turn producin Quenya an?o, an?u&.) The noun lan?o %throat% was li"ewise said to have the plural form lan?.i, probably implyin that Tol"ien intended lan?o to be derived from langu, but then he chan ed lan?o to lanco. Aee entries L,KD, L,NK in +tym. It may well be that the replacement form lanco is similarly to be derived from lan u, in which case it should have the stem lancu& and plural form lan9ui, thou h neither is attested. 5ot includin early %Qenya% material, this list should cover nearly all the "nown D&stems# R an?o (an?u&, pl. an?.i) %sna"e% (+tym entry ,N5:,, compounded an?u& in an?ul'c, +tym entry L9K) R l"co (evidently l"cu&) %wax%- compare the related word l"cu%a %taper, candle% R %alo (%alu&) %pollen, yellow powder% (primitive smalu, +tym entry !M,L) R orco %=rc, oblin% (orcu&, pl. or9ui) (primitive Rr u, +tym entry S.9K, or a Aindarin&influenced form of urco FB.v.G, ./#039- this Aindarin&influenced form may also be treated as a normal noun in &o and thus have the pl. form orcor, compare 72#?8 and ./#039) R ranco (rancu&, pl. ran9ui) %arm% (primitive ran u, +tym entry .,K) R rusco (ruscu&, pl. rus9ui) %fox% ('7#0IH, CT84#49) R s1lo (s1lu&) % oblet% (primitive suglu, +tym entry !1K) R ulco ( ulcu&, pl. presumably ul9ui) %support, prop% (primitive tul u, +tym entry "1L1K) R u%,o ( u%,u&) %deep valley (under or amon hills)% (primitive tum#u, +tym entry "1%) R u%2o ( u%2u&) %hump% (the entry&head "1M$1 in +tym apparently represents a primitive word, not Must a simple %root%) R un!o ( un!u&) %hill, mound% (primitive tundu, +tym entry "1N) R ulco %evil% as noun (ulcu&- the ablative ulcullo is attested, CT80#4H, H0&H8- pl. probably ul9ui) R urco %bo ey- =rc% (urcu&, pl. ur9ui) (primitive ur u or uru u, ./#039)

R U u%no (U u%nu&) %Dtumno%, 7el"orLs first stron hold (primitive form either 1tupnu, 72#>3, or 1tu#nu, +tym entry "1%) In the case of rauco %demon%, Tol"ien seems to be uncertain whether the primitive form was rau u or rau 2, or maybe he meant both to have co&existed (./#039). If it was rau u, then Quenya rauco should have the stem raucu& (and plural rau9ui;) =ne plural form of the compound Valarauco %)emon of 7i ht% (./#84I, Aindarin #alrog) actually occurs in the !ilmarillion, but its form is most surprisin # we have Valaraucar with the vowel a before the plural endin O This must reflect an alternative and Buite distinct primitive formation rau &. (Aee %Cariation of final vowels;% below.) Tuo %muscle%, which Tol"ien derived from primitive tTgu (entry "15 in +tym) may be a peculiar case. If uo is to be a D&stem, it mi ht conceivably assume the form 1& (for tU+u&) before endin s, or Must u& before a consonant cluster (e. . dative 1n %for a muscle%, ablative ullo %from a muscle%). But it is also possible that u+u would under o dissimilation to uo (as o+o is "nown to do) rather than bein contracted to 1- if so, we would simply see uo& also before endin s. =ne probable feature of the I0 and 1&stems is nowhere discussed in published material, but seems to be a necessary conseBuence of the eneral system# In the archaic %Boo" Quenya% that preserved a distinct %obMect% or accusative form, li"e ciry) as the accusative of cirya, the len thened final vowels would presumably preserve the ori inal Buality of the vowel# It was only short ori inal final &i and &u that turned into & and &o in Quenya. Thus $in %hair% with stem $ini& would presumably have the accusative form $in", whereas s1lo % oblet% with stem s1lu& would have the accusative form s1l1. But in Quenya as spo"en in 7iddle&earth, the distinction between nominative and accusative was abandoned- %it was adeBuately expressed by word order% ('lot! letter). 'honolo ically, the older accusatives $in" and s1l1 could have produced +xilic Quenya $ini and s1lu, still remainin distinct from nominative $in, s1lo 6 but Tol"ien appears to be tellin us that the distinct accusative forms were abandoned alto ether. Mariation of final vowels( Net another "ind of stem variation is so poorly attested that we cannot be sure whether it is really meant to exist %internally%, or merely reflects Tol"ienLs indecision as to the exact shape of a word. .e have already mentioned how Valarauco %)emon of 7i ht% (./#84I, Aindarin #alrog) has the plural form Valaraucar, if we are to believe the !ilmarillion. .hy this shift from a final &o in the simplex form to &a& before the plural endin &r; .ould we see the same chan e before other endin s, e. . dative ;Valaraucan : %for a Calarauco%; .e have only one other possible example of this variation of &o vs. &a&# In the Etymologies, the Quenya word for %root% is sun!o (entry !1/), and this word

also occurs elsewhere with reference to linguistic %roots% or bases (./#043). Net the plural form %roots% appears to be sun!ar in the compound Tar%asun!ar %2oots of the 'illar% (the name of the slopes of 7t. 7eneltarma in 5Pmenor, DT#4>>). But such variation of &o vs. &a& seems difficult to Mustify within Tol"ienLs scenario for how Quenya had evolved from primitive +lvish. (By contrast, it is easily explained why final &o and & sometimes turn into &u& and &i& before endin s# ori inal short &u and &i were chan ed in final position, but not elsewhere.) It should be noted that in neither of the two %attested% cases are the sin ular and the plural form provided in the same source, or even "nown to be closely contemporaneous. Ao maybe Tol"ien sometimes thou ht of the Quenya word for %Balro % as Valarauca (hence pl. Valaraucar) rather than Valarauco, and maybe he li"ewise wavered between sun!o and sun!a as the word for %root%. It is also possible that sun!o came to be used especially with reference to a lin uistic %root% (base, root&word), whereas sun!a refers to a natural %root% (of trees etc., also used metaphorically of the slopes of a mountain). Lost final vowels preserved #efore 7consonant8 endings( .hen an endin be innin in a consonant (such as &nna for allative) is to be added to a noun endin in a consonant, a connecting vowel is slipped in before the endin to avoid an impossible consonant cluster (unless the endin itself is somehow simplified). =ne may use &e& as such a connectin vowel, this bein the most neutral vowel (as in the attested allative form Elen!ilenna %to +lendil%, '7#894). Quenya lost certain short final vowel as the lan ua e evolved from primitive +lvish, and one could speculate whether these vowels would be preserved before certain endin s. As noted above, the name *an!os has the stem *an!os & because the final element is a reduced form of os o %castle, fortress- city% (72#0I9). Ahould the ori inal final &o of os o be preserved in some case forms, e. . ; *an!os onna as the allative %to 7andos%; =r would the neutral connectin vowel &e& be introduced here as well, so that we would have ;*an!os enna instead; .hat about a noun li"e ol %island%; It is said to be derived from primitive tollo and therefore has the stem oll& as in the plural form olli (typo %tolle% in +tym, entry "9LI). But would the lost final &o of the ori inal tollo be preserved in some cases, e. . locative ; olloss %on an island%; If so, the stem of ol would have to be cited as oll(o)&, the final o turnin up only before endin s be innin in a consonant. The instrumental form of a%,ar, a%,ar & %doom, fate% (more commonly u%,ar, u%,ar &) is attested as a%,ar anen in DT#401. 5otice the vowel &a& turnin up before the instrumental endin &nen. Tol"ienLs idea may be that the word for %fate% was somethin li"e m+#arta in primitive +lvish, and that the ori inal final &a is preserved in the instrumental form a%,ar anen, thou h this vowel (as well as the & precedin it) has been lost in the simplex form a%,ar %doom%.

The locative form of )l, al& %foot% is said to appear as alass in one late manuscript (CT80#4>- I assume that the )l here discussed means %foot%, thou h no loss is provided in this source). A ain the idea may be that %foot% was tala in primitive +lvish, and that the ori inal final &a is preserved before endin s be innin in a consonant. Net the material is 6 as usual 6 less than consistent. In the Etymologies, entry ",L, the enitive (later presumably dative) form of )l is iven as alen, which form is probably best interpreted as al& plus the enitive (later# dative) endin &n, the normal connectin vowel &e& materiali!in between the noun and the endin to avoid the impossible form ** aln. .ould Tol"ien later have said that the dative form of )l, al& should be alan rather than alen, to o with alass as the locative form; The form alass is supposedly ta"en from a fuller declension of )l, so the same (ca. 43>?) source document may provide the answer 6 but it is not available to us. 5o definite conclusions can be reached at this sta e- I would normally use the %neutral% connectin vowel &e& unless there is stron reason to believe that another vowel is to be preferred. It should be noted that &i (rather than &r) is in any case the preferred plural endin , so even if ol, oll& %island% does appear as ollo& before endin s be innin in a consonant, the plural %islands% should be olli (attested in (T4#1I) rather than ** ollor. The same principle would apply to all case endin s incorporatin the plural element &i&, li"e the plural dative &in or the plural instrumental &inen, and it would also apply to the other case endin be innin in a vowel, the enitive in &o. Ao even thou h %by doom, by fate% is attested as a%,ar anen, a correspondin plural form %by (the) fates% should probably be a%,ar inen with no &a& before the endin . The enitive would have to be a%,ar o (for older am#artao would be so altered even if it once did exist). .=2)A .+ )=5LT QDIT+ U5=. $=. T= I5@(+,T There are certain Quenya nouns and verbs that by their shape are difficult to inflect in the various rammatical forms. Applyin the normal rules may result in sli htly weird, or even phonolo ically impossible, forms. .e will survey some of the problematic roups of words. N91N! As noted above, there are some uncertainties relatin to the behaviour of nouns in &il derived from primitive &la. $ere follow certain other cate ories of %stran e% nouns. D? Monosylla#ic nouns with a long vowel? The vowel of these words must presumably be modified in many of the inflected words. .e will list the words involved. .ith the vowel # %) %hand% and r) %lion%. Aome more occur in early %Qenya%, e. . c) %Maw%, r) %arm%, s) %fire%, .) %wind% (but other, polysyllabic words for these concepts are attested in Tol"ienLs later Quenya). =ther %Qenya% words of

this "ind include ) %mind, inner thou ht%, c) %deed%, $) %lower airs%, l) %open space, moor% and 9u) %duc"% (but ) and l) have other meanin s in later Quenya). .ith the vowel O# 23 %lip% (so in a late source published in CT03#3- in the Etymologies the same word is lossed %mouth%), r3 %day% (a full cycle of H8 hours- %day% meanin the hours of dayli ht is aur). In early %Qenya% we also have $3 %last hour, death%, ny3 %a bleat%, s3 %eyeball%, y3 %tea%, V3 name of a Cala. .ith the vowel *# n" %woman- a female% (stated to be a poetic word- compare the more re ular word n"s, niss&), 2" %a fly, small insect% (but %spec", spot% in early %Qenya%). .ith the vowel R# l' %ni ht, a ni ht%, n' %conception (: idea)%, ' %wool%. In %Qenya% we also have forms li"e #', y', both meanin %son% (but these forms Tol"ien apparently replaced by yon!o later). The %Qenya% form B, a poetic word for the sea, may also be mentioned, as may 0' %owl%. .ith the vowel U# c1 %arch, crescent% (in early %Qenya% also %crescent moon%), l1 %a time, occasion%, R1 %)rW , member of a certain 7annish race of 7iddle& earth% (see DT#01I), s1 %noise of wind%. The latter also occurs in early %Qenya%, which lan ua e also has r1 %steadfastness% (hardly a valid word in (ot2&style Quenya). =ne Buestion without any entirely clear answer is how to form the nominative plural of these words. 5ormally, words endin in a vowel (except &) add the endin &r in the plural. )oes this also apply to words li"e these; If c1 is %crescent%, should %crescents% be c1r; Aome information about the word %) %hand% surfaced in CT8?#>. Tol"ien stated that it never received the plural endin &r, in part because this would ma"e the word clash with the noun %)r %dwellin % (instead, only the plural endin &li was used# %)li %hands% 6 thou h otherwise, the endin &li is associated with the special %partitive plural% discussed above). Aince the form **%)r was not used for such a specific reason (the clash with %)r %dwellin %), this may indirectly su est that most of these nouns can indeed have plural forms in &r. .hat about the nouns in &3; .e "now that polysylla#ic nouns in & normally form their plurals in &i. Ahould the plural form of r3 %day% be r"; =r maybe r3r; @or what it is worth# .hen the word r3 %day% appears at the end of a compound, it is shortened to &r, with a plural form in &ri (compare :es ar, 5ew NearLs )ay in the +lvish calendar, with en!eri %middle&days%, days inserted between certain months). But this may not necessarily support ;r" as the plural form

%days% when the word occurs by itself. At the end of compounds, words that normally have plural forms in &r may be transferred to the i&plurals instead. @or instance, the prominent r&plural Nol!or (archaic N?ol!or) may be contrasted with &n?ol!i in the compound E yan?ol!i %+xiled 5oldor% (./#0?8- the sin ular is probably E yan?ol with stem E yan?ol!&). .ords may be shortened and reduced at the end of compounds, li"e r3 loses its lon vowel in the compound form &r, and this may also have conseBuences for the way the word is inflected. 5ot all of these words behave in such a way, however. In the case of the n' %conception% (idea), Tol"ien indicated that this word has the stem&form n'.& (he listed the pl. n'.i), preservin the second consonant of the ori inal root N9:9 listed in the Etymologies. The word ' %wool%, derived from a root "9:, may li"ewise have the stem '.&. Before the enitive endin &o, this . would li"ely chan e to #, since .o is apparently not allowed in Quenya. If so, the enitive of n' and ' would be n'#o, '#o. The word r) %lion% has the stem&form r)#&, hence pl. r)#i (listed by Tol"ien) and enitive presumably r)#o. But it is not always clear how the enitive endin &o would be added to these words. Addin the enitive endin &o to a word li"e l' %ni ht% would presumably not alter the word in any way- the &o simply mer es into the already& lon ', and only the context would indicate that the noun is to be understood as a enitive. But what about the words in &); .e "now that the enitive endin &o normally displaces a final &a, as in Var!o %CardaLs% (Namri). Ahould the enitive form of %) %hand% be %'; =r could it simply be %)o; The combination ao is apparently not allowed in Quenya- does it ma"e any difference if the ) is lon ; 6 In the case of monosyllabic nouns endin in other vowels than &) and &', we must probably assume that the enitive endin is simply added to the word. The possessive endin &#a would also be added directly, presumably mer in with &# and &. in the case of the nouns with stem&forms in these consonants. %=f a lion% should evidently be r)#a- one may wonder whether n'.a or n'#a is the best option as the possessive form of n', n'.& %conception%. Before endin s with a double consonant or a consonant cluster, the lon vowel of these nouns must be shortened. Thus the plural ablative of %) %hand% is attested in (2#?H as %annar (since %into hands% could not be the impossible form %)nnar). .e must assume it would be the same with pronominal endin s, e. . lul%a %our time% from l1 %time% (hardly l1l%a). $owever, the example %)rya %her hands% from Namri indicates that the lon vowels of these words do remain lon before pronominal endin s includin combinations in &y&, such as &nya %my%, &lya %thy% and &rya %hisQher%. As we discussed in (esson @ifteen, it is not entirely clear how we are to analy!e roups li"e ny, ly, ry# are they

single palatali!ed consonants, so that a lon vowel can survive in front of them; But they seem to count as consonant clusters, or lon consonants, for the purpose of determinin which syllable receives the stress. H. Nouns in 0i and 0i= There are also some unanswered Buestions relatin to nouns in &i and &i, li"e li %people%, arani %"in dom% or )ri %Bueen%. $ow is the enitive endin &o to be added to nouns in &i; There would be three vowels in seBuence, and words li"e lio, aranio would have to be accented on the i, soundin rather aw"ward. Aome believe the would drop out, producin forms li"e lio or aranio, but this cannot be proved. The present writer is inclined to thin" that the e would be len thened to 3 so as to attract the stress# li3o, arani3o. .hat about the case endin s already includin the vowel &i, notably &in for plural dative, &i#a for plural possessive and &inen for plural instrumental; (There is also the %mystery case% from the 'lot! (etter, probably a short locative, that has the endin &is in the plural.) As far as we "now, the plural dative of )ri %Bueen% must be formed by combinin )ri and &in, but what does this result in; )o we see )r"n with a lon ", for tri0in; In the 'lot! (etter, Tol"ien states that in words of two or more syllables, a lon vowel was shortened before a final consonant. If so, older )r"n would become )rin, identical to the singular form of the dative ( )ri J &n). It may be that only the context can tell us whether )rin is sin ular %for a Bueen% or plural %for Bueens%. =n the other hand, in the possessive case there may be a distinction between plural )r"#a %BueensL, of Bueens% ( )ri J &i#a) and sin ular )ri#a %BueenLs, of a Bueen% (simply )ri J &#a). The forms would be stressed differently, ma"in them clearly distinct. .ords in &i (mostly abstracts, erunds and some feminines li"e Vali) pose their own problems in this re ard. .hat is the dative plural and possessive plural of, say, Vali; Addin the endin s &in and &i#a produces the impossible forms **Valiein and **Valiei#a# Quenya does not possess the diphthon ei. .here it formerly occurred it normally became ", but accordin to a note by Tol"ien published in CT81#?, it turned into 3 followin short i. If so, impossible *Valiein first turns into Vali3n. Then, accordin to the 'lot! rule Must mentioned, a lon vowel before a final consonant is shortened in polysyllabic words# Thus we presumably arrive at Valien. As in the case of )rin, the distinction between sin ular and plural is lost, for Valien could Must as well be the singular dative %for a CaliX% (Vali J &n). Aimilarly, plural possessive **Valiei#a would turn into Vali3#a, which could also be sin ular instead (as if Vali J &#a). The instrumental forms would be similar to the possessive forms deduced above, only with &nen instead of &#a.

It is almost temptin to start from the nominative plural Valier and use ; Vali3rin as the dative plural (and perhaps ;Valier.a as the possessive pluralO) But there is no trace of such forms in published Tol"ien material. 0. Nouns in 0oa# Quenya possesses a number of nouns endin in the combination &oa, most notably coa %house%, 0roa %body%, loa %year% (literally % rowth%), noa %conception% (idea), roa %do %, oa %wood%. 7ost of these are meant to be descended from earlier forms in &awa or &aw (see especially CT8?#0I), a combination that became &oa in Quenya. In early %Qenya% we also have $oa %hoard, treasure%, loa %life% (probably obsoleted by loa %year, rowth% in later Quenya), %oa %sheep%, oa %wool%, 2oa %beard% and roa %wild beast%. *oa, oa and 2oa are perhaps obsoleted by %)%a, ' and $an?a of the same meanin in Tol"ienLs later Quenya- moreover, roa was redefined as %do % (see above). Anyway, it seems that most of these were also meant to represent older forms in &awa or &ava, althou h $oa and oa are derived from roots where the first vowel was o from the be innin . (The same oes for the later form noa %conception%, derived from a root N9:9 in the Etymologies.) The main %problem% relatin to these forms is this# .hat happens if we add the enitive endin &o; Aince this endin displaces a final &a, what does a word li"e 0roa %body% turn into in the enitive case; Is %bodyLs% to be 0r', for the impossible form 0ro7o;O If we are to ta"e into account the supposed phonolo ical development, we could reach pretty surprisin conclusions. The word 0roa is meant to be descended from primitive sraw, srawa (72#0I9, CT8?#0I). The enitive endin &o represents primitive &ho, mer in into a final &a after the loss of h. If srawa0ho became srawao and then sraw2 early enou h, the Quenya outcome would be 0ra#oO Aimilarly, the enitive of coa %house% would be ca#o, the enitive of loa %year% would be la#o, and so on (reflectin ancient forms li"e awa %house%, lawa % rowth, year%# CT8?#0I, 8H#49). $owever, the enitive of noa %conception% is presumably no#o, since the root is N9:9 rather than **N,:,. (,ompare n'.i as the plural form of n', a shorter synonym of noa derived from the same root. Before &o, we must assume that . becomes #.) ME.%! In the case of a few verbs, it is difficult to predict the form of the gerund and the perfect tense. =ne such verb is $euya& %abhor%, derived from the root $-E: (see this entry in the Etymologies). Accordin to our eneral understandin , the

erundial endin &i displaces the endin &ya. But %$eui% would be a rather unli"ely form as the erund %abhorrin %. 'erhaps we must refer bac" to the ori inal root $-E:. In $euya&, the ori inal w has chan ed to u before another consonant# the y of the verbal endin &ya. But between vowels, the w of 'rimitive +lvish normally turns into Quenya # (the only notable exception seems to be that . is preserved unchan ed followin the diphthon ai, as in ai. %bird%). Ao $-E: %abhor% J the erundial endin &i would perhaps come out in Quenya as $e#i. Aimilarly, the perfect %has abhorred% could conceivably be e$3#i. S1ya& %to breathe% is even more problematic. Ahould the erund be sui, the first vowel of the erundial endin &i mer in with the stem&vowel to produce the diphthon ui; And will usui do as the perfect %has breathed%; Cerbs in 3 J ya would be particularly difficult. 5o such verb seems to be attested, but they probably could exist# In the Etymologies we have thio as the 5oldorin (later# Aindarin) verb %to seem%. The root is "-V, so the primitive form is clearly meant to be thJy&- this would apparently come out as s3ya& in Quenya. Aome post&Tol"ien writers have already used this unattested verb. <ranted that this extrapolation is correct, what would the erund of such a verb be; Aince Quenya does not have ei, we cannot o for **sei. $istorically, older ei normally becomes lon " in Quenya. S", then; 'erfect es", the au ment reflectin the ori inal stem&vowel of the root "-V; The verb %to chan e% is apparently a0ya& (only attested in the past tense# a0yan). Ahould we o for a0i as the erund %chan in %; The problem is that 0y is not 0 J y, but a sin le unitary sound, li"e the <erman ich0Laut (of which the h of +n lish huge, human is a wea" version). In a0i the 0 would be a different sound, li"e normal +n lish breath&0. Ahould we use a0yi, then; $ardly- Quenya phonolo y does not seem to allow 0y before the vowel i. It is a problem that we donLt "now what the ancestral form of a0ya& is supposed to be# maybe a h0y; If the ori inal root is ,K-, one could ar ue that the erund should indeed be a0i# =ri inal h becomes 0 between vowels, whereas h J y would produce 0y, as in a0ya&. The perfect %has chan ed% could conceivably be a0)0i, if we are ri ht to assume that the entire initial syllable is prefixed as an au ment when the verb be ins in a vowel# primitive a h0 hiiJ. But there are also those who would ar ue that a form li"e a0y)ni may be used, based on the past tense a0yan. Tol"ien indicated that %intrusion of n from the past FtenseG% into the perfect tense sometimes occurred in the lan ua e (./#0>>). +arly %Qenya% has verbs in &i ya, li"e nai ya& %to shame, abuse% and 2ai ya& %to repay%. If it is at all possible to fit them into Tol"ienLs later scheme, we must assume that the erundial endin &i replaces &ya (so that we would have nai i

%shamin %, 2ai i %repayin %), and then maybe anai i, a2ai i in the perfect tense. The verb uia& %to sprin , sprout% is apparently derived from the root "16 with the simple endin &a, early tuya& becomin uia& in later Quenya. 7aybe we could have the erund ui, for impossible tuyi, and the perfect u ui, for utUyi;O A verb li"e lia& %twine% (from early %Qenya%) is very difficult to adopt into Tol"ienLs later inflection scheme. It is said to be derived from liya (Q(#I0). A erund li (for impossible liyi) would clash with li %people%, a noun that was present already in Tol"ienLs early development of QenyaQQuenya (it is indeed listed on the same pa e in the Q( as the verb lia&). The perfect, if it is at all possible to construct one, would have to be somethin li"e il" (for il*yi). .e can hardly avoid the problems encountered when we try to construct the perfect tense forms of such cumbersome verbs, but the gerund in &i may perhaps be replaced by some other abstract endin - the best alternative is probably &l. Ao from verbs li"e a0ya& %chan e%, s1ya& %breathe% and lia& %twine%, we could derive abstract nouns li"e a0yal %chan in , chan e%, s1yal %breathin % and lial %twinin %.