Analogical leveling in Quenya compounds (and prefixed words)

Abbreviations used:

WJ: The War of the Jewels
LR: The Lost Road
MR: Morgoth's Ring
PM: The Peoples of Middle-Earth
LT: The Lost Tales
LOTR: The Lord of the Rings
SIL: The Silmarillion
UT: Unfinished Tales

VT: Vinyar Tengwar
PE: Parma Eldalamberon

Entries in the Etymologies are referenced by their CE roots in this article, all references to 'Appendix ???' refer to 'The Lord of the Rings'

As Helge Fauskanger put it in the appendix of his Quenya course,

seemingly "irregular" features of the language are very often justified by the long historical evolution Tolkien envisioned. For instance, when the noun talan "floor" has the plural form talami instead of talani, this is because the original Primitive Elvish base had the form TALAM: As the distinctive features of Quenya phonology evolved, final -m was no longer tolerated and was altered to the closest "permissible" sound: -n. Hence older talam appears as talan when the word occurs without endings. But when endings are added so that a vowel follows, the original -m was not final and therefore did not have to be changed. Hence the plural form talami "floors".

Consequently, there are a lot of things which could happen when a word is prefixed or the second element of a compound and hence its initial sound is not inside the word and would not undergo changes relevant for the initial sound. Indeed, in Sindarin this is probably the reason for the frequently encountered consonant mutations in compounds. To quote again from Helge Fauskanger's course:

Yet one could plausibly argue that the perfect tense of lanta- should actually be arantië!

So what about Quenya - is the scenario outlined above plausible? The fact that these changes could happen doesn't imply that they do happen - analogical leveling could just force the parts of compounds to be the same as unitary words, i.e. prevent changes from occuring. In order to understand if historical development or analogical leveling persists in compounds, we have to investigate all relevant compounds, i.e. all for which changes could potentially occur.

The -mar group:

The element -mar is derived from MBAR, as such the -mb- could preserved when this sound is placed inside a compound, see a-mbar derived by an extension of the root. However, in all compounds this doesn't ever happen:

Eldamar 'Elvenhome' (ÉLED; MR:176)
Endamar 'Middle-earth' (EN, MBAR, NDOR)
Fanyamar 'upper air' (SPAN)
Mittalmar 'Midlands' of Númenor (UT:165)
Noldomar 'Gnomeland' (LT1:262)
Valimar 'Vala-home' (Nam, RGEO:67)
Vinyamar *"New Dwelling" (SIL)

Note that there is an independent word mar, cf. Mar-nu-Falmar 'Home under Waves' (SIL) which can trigger the analogical leveling process.

The -mando group:

These words consist of the blended roots MANAD and MBAD and are hence protected from reverting to -mb- when compounded:

Angamando (SIL)
Calamando 'light Mando' (MBAD, KAL, MANAD)
Morimando 'dark Mando' (MBAD)

Note that there is likewise an independent word mando 'custody, safe keeping' (MR:350) which can trigger analogical leveling.

The -lant group:

The blending of DAT and TALÁT leads to a group of words which in compounds always shows analogical behaviour, the initial l- is always protected:

Atalantë 'the Fallen' (DAT/DANT, TALÁT, SIL, SD:247, 310; also LR:47)
lasselanta 'leaf-fall' (Appendix D, Letters:428)
Noldolantë 'the Fall of the Noldor' (SIL)

The -n(d)or(ë) group:

Derived from NDOR, we might expect that -ndore reappears in compounds although the full word nórë shows nd > n word initially. The group shows a large variety:

In two cases, confusion with nórë 'clan' from NÔ, ONO is explicitly identified as the reason for later analogical leveling:

Valandor 'the land of the Valar' (SA:dôr, Silm) became known as
Valinórë 'the people of the Valar' (SA:dôr, Silm)
Númendor 'land of the west' (SA:dôr) became known as
Númenor, full form Númenórë (LR:47, SD:247, NDÛ)

Following an vowel, consistently -nor(ë) is preserved:

Aiwenor, Aiwenorë 'Birdland' (AIWÊ)
Arnanórë, Arnanor 'Arnor' (Letters:428)
Eldanor 'Elvenland' (MR:176)
Firyanor 'Hildórien' (PHIR)
Lestanórë 'Doriath' (WJ:369)
sindanórië 'grey land; (LOTR)
Lómëanor 'Gloomyland' (LOTR)

Finally, especially following -n, probably intentional dissimilarity leads to -nd- and no analogical forms:

Andórë, Andor 'land of gift'(SD:247)
Endor 'Middle-earth' (SA:dôr, NDOR)
Endórë 'Middle-earth' (Appendix E)

The -ndil group:

In spite of NDIL yielding words like nilmë or nildo where we see nd > n word initial, words using this affix are usually not subject to analogical leveling but preserve -ndil. However, since there isn't actually an independent word *nil it doesn't seem reasonable to treat the group as compound words, perhaps one should rather interpret them as having an affix -ndil:.

Thus, when the first word ends in a vowel, the full affix -ndil appears:

Aiwëndil 'Radagast' (LOTR)
Eldandil 'Elf-friend (WJ:410)
mámandil '*sheep-friend' (UT:209)

When the first word ends in -n, the contraction -n-ndil > -ndil is straightforward (note that a contraction has to occur since three consonant clusters aren't allowed):

Elendil 'Star-friend' (WJ:410)
Siriondil '*Sirion-friend' (Appendix A)
Vorondil '*Faithful friend' (Appendix A)

Finally, when there is no straightforward contraction preserving -ndil, a contraction producing the shorter -dil occurs (i.e. neither the analogical -nil nor the historical -ndil appears) to prevent a three consonant cluster -rnd-:

Anardil '*Sun-friend' (Appendix A)

The -ndur group:

Similar to the -ndil group, this group seems to be derived from NDUR without an actual independent word. Consequently, there is no reason to expect analogical leveling, and in fact it is not seen:

Aulendur 'Servant of Aulë' (PM:366)
Cemendur '*Earth-servant' (Appendix A)
Elendur '*Star-servant' (Appendix A)
Pelendur '*Fence-servant' (Appendix A)
Valandur '*Vala-servant' (Appendix A)

Again, we find a shortening to -dur in a contraction to prevent a three consonant cluster:

Isildur '*Moon-servant' (Appendix A)

The DOMO/ DO3 derivatives:

The blending of these roots results in lómë 'night' and derivatives with the sense 'twilight'. Following the consonant -n (and possibly others) th original -d- is recovered in compounds:

tindómë 'starry twilight, starlit dusk' (DOMO, TIN, SA:tin)
tindómerel 'daughter of twilight (TIN, SEL-D, SA:tin)

However, following a vowel l- remains:

Taurelilómëa 'Forestmanyshadowed' (LOTR)
Tauremornalómë '*Forest (of) Black Night' (LOTR)

Strengthened forms:

CE roots can be strengthened by prefixing the stem vowel, likewise an intensifying prefix can strengthen a later word. The strengthening of the CE root places always places the initial sound inside the word and the historically justified form is the outcome. Strengthening by a prefix follows a peculiar pattern of analogical developments (blendings with A-, N- and um and brings nasals by analogy to forms which wouldn't have these from historical development, see amparka 'very dry' instead of *apparka (VT45:5).

Presumably a strengthened CE root represent:

ambar 'doom' (MBARAT), hence *AMBARAT
Turambar 'Master of Doom/Fate'(Appendix A, SA:tur, TUR, MBARAT)
#ascat- 'break asunder' (SD:310), hence *ASKAT

Note that SKAT leads to a different development with a prefix (as opposed to presumed CE root strengthening) as apparent from

terhat- 'break apart' (SKAT)

which doesn't come out as **tereskat- (the cluster -rsk- would be forbidden of course).

We presumably see strengthening by a prefix with fortified initial consonant similar to (VT45:5) in:

ingolë 'Science/Philosophy' (PM:360)
ingólemo 'one with very great knowledge' (PM:360)
ingolmo 'loremaster' (WJ:383)
Ingolondë 'Land of the Gnomes' (ÑGOLOD)
andúnë 'sunset' (NDU)

Presumably these strengthened forms lead by analogy to

Lambengolmor 'Loremasters of Tongues' (WJ:396)

where an unusual historical development crops up (the prefixed words ends in a vowel), but cf.

mólanoldorin 'the language of the Noldor enslaved by Morgoth' (MÔ)

where analogical leveling occurs for the same root.

Various compounds and prefixed forms:

In most known prefixed forms and other compounds analogical development persists:

alahasta 'unmarred' (MR:254)

We don't know the root - if it is close to SKAT or SKAR, then it is instructive to see that we do not find **alaskasta, we then would see an analogical form instead of a historical. However, it is possible that the form is derived from a root with *KH- , in this case we don't learn much.

andamunda 'elephant' (MBUD)

Here we have direct evidence of analogical replacement since the CE compound andambundâ is given, proving that the word historically was a compound but the development proceeded for both elements separately, leading to mb > m inside the word.

Envinyatar 'Renewer' (LOTR)

Here, the prefix doesn't lead to the reappearance of -w- like *enwinyatar, instead the cluster nv which is not part of the allowed Quenya clusters in unitary words is allowed to stand to preserve the shape of the element vinya.

ilfirin "immortal" (PHIR)

Following the prefix, -ph- > -p- could occur, and indeed such a form ilpirin is listed by Tolkien but denoted as obsolete - the valid for is the analogical ilfirin.

Sangahyando 'Throng-cleaver' (Letters:425)

From pirya, CE: pisjâ we can determine that -sj- inside a word would develop into -ry-. However, in this compound, clearly the word initial development sj- > hy- is preserved by analogical leveling.

However, in a few cases, historical development occurs:

lindornëa 'having many oak-trees' (DÓRON, LI)

Here initial D- in the CE stem lead to n- like in norno 'oak tree', but in the compound version the original -d- reappears. This is in fact pretty consistent with the reappearance of -d- following a consonant which lead to initial l-, cf. the lómë group above.

Russandol 'Copper-top' (PM:354)

In this name, we presumably see an abbreviated version of nóla 'round head, knoll' (NDOL). However, as in the case of nildo/-ndil we observe that if a compound isn't made with the full word but with an abbreviated affix, analogical leveling isn't likely to occur and we see historical development preserving -nd- inside the compound instead, in spite of the prefixed word ending in a vowel.

Norrívë 'December' (PM:135)

The last element is clearly hrívë 'winter'. Since the first element of the compound ends in a consonant, analogical leveling cannot persist (a form like **Norhrívë would clearly be nonsense) and a contraction is made. However, cf.

ohlon 'diphthong' (VT:39:9)

where a simple vocalic prefix allows the analogical -hl- to stand as (unusual) two consonant cluster.


For the most part, the problem seems well understood. The main rule is: Exceptions from the main rules are: Thus, a prefix or prefixed word ending in a vowel leads almost always to analogical leveling, which makes a very strong case against the idea that the perfect augment would be capable of triggering historical development and produce a form like ?arantië from lanta-. There is only one exception in the above list, Lambengolmo, and that can be understood by analogy with the group around the strengthened ingolë. The strengthening of a CE root is a different mechanism from a Quenya prefix/compound and must not be treated on the same footing. Thus, apart from a few cases, the difficult scenario of historically justified forms reappearing with prefixes, augments and in compounds does not seem to be realized in Quenya.


I would like to thank Helge Fauskanger for interesting discussions leading to this article.

Thorsten Renk

Back to the index