A rogue's guide to Sindarin word reconstruction

Abbreviations used:

PQ: Primitive Quendian
CE: Common Eldarin
S: Sindarin
N: Noldorin
Q: Quenya
Why is this article called 'rogue's guide'? Because I feel that word reconstruction is never a good thing to do. Why is this so? Because in reconstructig words, we leave the rather clear path of what is attested and enter in speculation, and here, the dividing line between what is sensible to do and what not soon becomes murky. Consider a simple example, the verb 'to be' in Sindarin. In LR:374, we find the entry

NA- cf. ANA- be. Stem of verb 'to be' in Q.

Now, Tolkien could have easily written stem of verb 'to be' and left out the Quenya part, but in fact he did not. So there is a fair possibility that he wanted to imply that this is the stem of the verb 'to be' in Quenya, but not in Sindarin. This is not far-fetched, in LR:348, entry AM-, we find an explicit example of a root which is not relevant for Sindarin. Thus, if we start from a Quenya verb and construct its Sindarin equivalent, we simply do not know if we do something Tolkien would have done or not.

Bluntly said, reconstructing words and using reconstructed words essentially means leaving Sindarin and writing in Neo-Sindarin. It is sometimes argued that Tolkien would have constructed a word in a similar way if he would need it, but the truth remains that word reconstruction is by no means an unique procedure and that we simply do not know if he would have accepted the outcome. There would also be the practical problem that a text needs to be understood by the reader. If the reader is not able to figure out what an Elvish text means by his own knowledge of Elvish but needs lengthy explanations about what reconstructions have been done, then the text probably contains too many of them.

Why then do I bother to write this guide? Because even if word reconstruction is a bad thing, there are situations where I feel the alternative is even worse, and if one is in such a situation, one should at least know to do the reconstruction properly. In general, I try to reformulate sentences such that no reconstructed words are necessary. But occasionally, a 100 line poem hinges on one missing word and any paraphrase would be extremely awkward - in this situation, a well-done reconstruction is clearly superior.

1. General outline

In talking about word reconstruction in the following, we do not refer to the rather obvious possibility of form Sindarin words by means of compound formation and prefixes. For example, in order to say 'to see again', we may feel free to use the prefix ad- (again) with the verb cen- 'to see' to form *achen (to see again), and there is really no need to assume one has to go back to the CE roots of both elements to form this word.

What is meant in the following is the possibility of add words to the vocabulary which (in the internal history) would not reflect modern compounds but really words along with the rest of the corpus. To give an example, Quenya has a word for 'to lie down', Q: caita-, Sindarin has no such attested words. However, using known rules of phonetic shifts (their discussion is the main purpose of this article), we can construct S: caeda- which is evidently not a compound word.

In order to do this type of word reconstruction, we must understand the development of the Elvish languages in time, not just their 'final' status. Typically, reconstructions for Sindarin are a three phase process.

a) Starting from a CE root, word derivational elements are used to form verbs, adjectives and nouns from this bare stem

b) The resulting CE word then undergoes changes towards the Old Sindarin stage. Here, certain vowel and consonant shifts are likely to occur.

c) Finally, coming from Old Sindarin to modern Sindarin, yet more changes occur; final vowels are lost and the well-known mutations and i-affection appear.

We will discuss these three stages in some detail, although not in its full complexity and give examples for both attested words and reconstructions in the following.

In particular, we will carry out the reconstruction procedure for the following examples:

Example 1: We are looking for a verb corresponding to Q: caita- (to lie down) in Sindarin
Example 2: We are looking for a verb 'to learn' in Sindarin
Example 3: We would like to construct a word 'privilege'
Example 4: We seek a word for 'seduce'
Example 5: We would like to construct a word for 'saddle'

Part I: The CE stage

In this part of the article, we will investigate how a word in CE is formed from the stem and what modifications the stem can undergo in the course of this process. The treatment here is far from a complete list of all possibilities and is intended to show the most important of those developments.

1. Stems and their modification

Stems or primitive roots are the basic building blocks of Elvish word derivation. While they themselves are as such not words in any Elvish language, any word in Elvish seems to be formed from a (modified) stem with the optional addition of a derivational suffix. Tolkien himself calls the theory of stem modification sundokarme. Let us start by characterizing the different types of stems:

Each stem consists of a consonant base (sundo) and a determining vowel (sundóma). According to the number of consonants involved, we may distinguish three different types of stems (note that the words given after each stem in the following are not a translation of the stem; this cannot be done because a stem itself is usually no valid word, but rather a description of its meaning):

a) 1-consonant stems, e.g. ES- (indicate, name)
b) 2-consonant stems, e.g. NAK (bite)
c) 3-consonant stems, e.g. GARAT (fort, fortress)

Note that the two vowels in 3-consonant stems are usually the same determining sundóma. Furthermore, we count stems like KWAT (full) as a 2-consonate stem, because the cluster KW is not broken up by a vowel and really represents one sound. The stem vowel can optionally be repeated after the last consonant, i.e. ES and ESE would be two representations of essentially the same object.

Selecting a stem basically determines the 'topic' of all derived words. However, a stem can be modified further to describe variations of the topic. There are several possibilities to do so:

a) Using the stem vowel as intensifying prefix:

Repeating and prefixing the stem vowel usually leads to a more intense version of the basic theme of the stem, hence LAK (swift) but ALAK (rushing); NAR (fire) but ANAR (sun); THIL (shine silvery) but ITHIL (moon), LED (go, travel) but ELED (depart, leave). Sometimes, if the stem vowel is prefixed and postfixed, it may vanish from the middle position, cf. RUK (to fear) with variant URKU (via URUKU).

b) A-infixion as intensifying modification:

Alternatively, an infixed A can act as an intensifying element. This is probably only possible if the stem vowel is I or U. For examples, see e.g. RUK (fear) but RAUK (very terrible creature); TUR (power) but TAUR (of vast power).

c) Fortification to nasalized stops

Stems beginning with B, D, G or M, N and Ñ can be fortified by turning them into MB, ND or ÑG. This fortification seems to correspond to an elaboration of the topic of the stem on a more abstract level, sometimes also denoting the consequences of the stem, see e.g. NAK (bite) and NDAK (slay), BAD (judge) and MBAD (duress, prison), DUL (hide, conceal) and NDUL (dark, obscure).

d) Fortification by repeating a consonant

Sometimes, a consonant inside the stem is repeated for an intensified meaning (this consonant can be brought into the stem by repeating the stem vowel), e.g. BAT (tread) and BATTA (trample).

e) Extension of the stem

Finally, there is also the possibility that a stem repeats the stem vowel (this is called ómataina and then ends in an additional consonant N, K, T, or S). This stem extension seems to denote again a result of the basic topic or a more abstract development, e.g. OR (rise) and OROT (height, mountain), KIR (cut) and KIRIK (sickle).

2. Derivational suffixes

After we have specified the topic of the word we want to construct and decided if we wish to use an intensitied, fortified or extended version of this stem, we have to decide on the word type. We can do this (and add further nuances to the meaning) by selecting suitable derivational suffixes. In PQ, we add these suffixes directly to the stem, regardless if any awkward sounds are produced. In the next section we will discuss the changes when we go to CE words.

1) Verb derivation

Verbs occur either as stem verbs (with no suffix at all) or as derived verbs, the second class is more numerous and hence more promising for the formation of new verbs. As derivational suffixes, we find

a) -tâ:
This suffix has a strong tendency to denote causative and transitive verbs (verbs describing an action on something). We find e.g. tultâ (make come, cause to come; summon, fetch) from TUL (come), maktâ (fight, wield a weapon) from MAK (sword), k'riktâ (reap) from KIRIK (sickle) and so on.

b) -yâ:
Verbs derived using this suffix have a moderate tendency to be intransitive, i.e. describe actions not directed onto an object, e.g. ulyâ (it is raining) from UL(U) pour, flow or beryâ (to dare) from BER (valiant).

c) :
This is a rare derivational element and it yields verbs of the mixed conjugation in Sindarin. No particular shade of meaning seems to be implied.

2) Adjective derivation

a) -i
Most color adjectives are derived with this ending, cf. karani (red) from KARÁN.

b) -kâ (-ka)
An unspecified adjective ending, cf. lauka (warm) from LAW (warm)

c) -nâ
This is often used to form adjectives which might also come out as perfect passive participles, cf. skarnâ (wounded) from SKAR (rend, tear)

3) Noun derivation

There is a host of noun derivational suffixes, we will only discuss some of them here:

This suffix is often found to denote intangible, abstract things, cf. ñgôlê (science, philosophy) from ÑGOL (wise). Often, the stem vowel is lengthened if this suffix is added.

A second meaning of this ending is to denote substances, cf. kyelepê (silver) from KYELEP (silver).

b) -dô, -ô, -rô (male) and -dê, -ê, -rê (female)
Those are used to denote the person doing some action, cf. lindô (singer) from LIN (sing)

c) -la
This suffix simply denotes a thing or a person, cf. makla (sword) from MAK (sword).

d) -mâ
This suffix often denotes implements or tools, cf. takmâ (tool for fixing) from TAK (to fix) or sukmâ (drinking vessel) from SUK (drink). Sometimes, it describes just vaguely an object connected with the stem, cf. parmâ (book) from PAR (compose, put together).

e) -mê
This suffix primarily denotes verbal nouns, corresponding to the English '-ing', cf. yulmê (drinking) from YUL (drink) but often in a more abstract meaning, cf. tekmê (letter) from TEK (make a mark).

f) -sê
This ending is often seen to describe the result of the action given by the root, cf. khotsê (assembly) from KHOTH (gather)

(Note that there are more derivational suffixes in CE than those (probably most common ones) given here - a more detailed discussion can be found in Helge Fauskanger's essay Primitive Elvish

3. Sound changes

Once we have formed a word from stem and suffix, a few changes are in order:

Consonant shifts:

a) y changes into i before consonants, e.g. kaitâ from KAY + -tâ, w into u.

b) The combination bm is reversed, cf. PQ: labmê -> CE: lambê

c) PQ: sd changes into CE: zd and PQ: ds into CE: ts

d) Medial -h- is lost, cf. the ending PQ: -hô becoming CE: -ô

Vowel shifts:

a) the short final vowels -a, -e and -o are lost.

b) short final -i is changed to -e

4. Examples

Example 1: It is not hard to decompose Q: caita- into the stem KAY and the verb derivational suffix -tâ. In CE, this yields directly CE: kaitâ (to lie).

Example 2: A suitable starting point for a verb to learn would be the root ÑGOL (wise). Since 'to learn' is rather not connected with an intensified form 'exceedingly wise' or a more abstract result, we use the unmodified stem. 'To learn' doesn't seem to imply a causative or transitive meaning (unlike 'to teach' = 'to make know' ), so we reserve the ending -tâ for 'teach' and settle for the intransitive ending -yâ instead, resulting in CE: *ñgolyâ (to learn).

Example 3: The nearest root to 'privilege' would probably be DAB (give way, make room, allow, permit). However, this is not quite the implied meaning, so we do a root fortification denoting the result of allowing, yielding *NDAB (privilege). We decide for the noun ending to denote intangible things, resulting in CE: *ndabê (privilege).

Example 4: 'to seduce' has definitely to do with 'cause to love', so we select the causative verbal suffix -tâ with some confidence to construct CE: *meltâ.

Example 5: 'thing having to do with horse' might not be the best possible description of a saddle, but let us select this form for the sake of the argument (it will lead to interesting developments later on) and pick -la as suffix, resulting in PQ: *rokla.

Part II: The OS stage

In the development from CE to OS, both vowels and some consonant groups undergo prominent sound shifts. Here, we try to indicate the most frequent of those.

1. Vowel shifts

The short final vowels -a, -e and -o are already lost in the CE stage, and those that are not will be when we discuss the shifts to Sindarin, so for the purpose of word reconstruction it is perhaps not really important to keep track of them. However, a PQ final short -u seems to become -o.

When not final, short vowels stay unchanged.

Long final vowels are shortened in OS, cf. CE: bélekâ -> OS beleka (mighty).

For long non-final vowels, the most drastic changes occur. Here, we find

â -> ó
ê -> í
ô -> ú

and without change

î -> í
û -> ú
For examples, see e.g. CE: ndâkô -> OS: ndóko (warrior, soldier) or CE: *rômâ -> OS: rúma

In Tolkien's later conception, however, rather

â -> å
seems to be relevant, this sound later evolving into au or o in Sindarin, e.g. CE: nâbâ -> OS: nåv -> nauv > S: naw (hollow).

Some attested diphthong changes are

ai -> ai
ay -> ae
ew -> eu -> iu
euy -> iui -> ui

For examples, see e.g. CE: beuyâ -> OS: buia

2. Consonant shifts

A prominent (and early) shift at the OS stage is kw -> p (this becomes q in Quenya instead), see e.g. CE: *kwantâ -> OS: panta (full). This is mentioned in the beginning, because many of the following shifts have to do with p,t or k and the conversion kw -> p has to be carried out first.

A large group of possible shifts acts on the consonants p, t and k. Here, in essence the changes of the stop mutation have to be carried out, i.e.

kk -> kkh
pp -> pph
tt -> tth

(those clusters will later become ch, ph and th in Sindarin) and likewise the changes of the liquid mutation, i.e.

lk -> lkh
lp -> lph
lt -> lth

rk -> rkh
rp -> rph
rt -> rth

In addition, we see some of the nasal mutation:

nt -> nth

Word initially, we find s- acting in a similar way (but not in medial position):

st -> sth (word initial)
sp -> sph (word initial)
sk -> skh (word initial)

(those clusters will also later become ch, ph and th in Sindarin)

For an example, see e.g. CE: alkwâ -> alpa -> OS: alpha (swan)

The above changes occur when k, p, t are trailing other consonants. However, those sounds are also changed when k, p, t occur before nasals. In these cases, the sounds are softened:

km -> gm
kn -> gn

pm -> bm
pn -> bn

tm -> dm
tn -> dn

See e.g. CE: yatmâ -> OS: yadme (bridge)

The (in Quenya) rather frequently occuring consonant groups ky, ty, ry, ny and syapparently shed the y at this stage. Presumably, word initial sy- becomes h- (for sure it does so in later Sindarin, so for our purpose, the distinction is rather academic).

ty -> t
ky -> k
ry -> r
ny -> n
sy -> s

sy -> h (word initial)

See e.g. CE: kyelepê -> OS: kelepe

If y is not part of such a cluster, it is turned into the full vowel i, cf. CE: *skalyâ -> OS: skhalia (veil, conceal). This is most important for the verbal ending -yâ.

Finally, various other assimilations are likely to occur:

bn -> mn
sm -> mm (medial only)
nm -> mm
dn -> nn
sr -> rrh
ln -> ll
ht -> tt
hs -> ss

3. Examples

Example 1: This yields CE: kaitâ -> OS: *kaita, the only change being the shortening of the final vowel.

Example 2: Here, we find CE: *ñgolyâ -> OS: *ngolia, the ending becoming short and the semi-vowel y shifting into i.

Example 3: For this example, we get CE: *ndabê -> OS: *ndabe, only the final vowel becoming short.

Example 4: Liquid mutation causes CE: *meltâ -> OS *meltha, again the shortening of the final vowel is found.

Example 5: CE: *rokla -> OS: *rokla remains essentially unaltered.

Part III: Towards Sindarin

In the development towards mature Sindarin, we encounter the same phenomena which make the life difficult for any Sindarin student: i-affection and lenition. In addition, some consonant groups show a behaviour that is not covered by the standard mutation charts.

1. Vowel development

Going towards modern Sindarin, a rich change in vowel structure can be observed. Arguably the most prominent effect is i-affection:

In the development from OS to Sindarin, the presence of the vowel i in a word tends to affect any other vowels inside the word. These vowel shifts should be known to the Sindarin student from the plural formation (if you do not know how to form the plural in Sindarin, this is not the right article for you now). Indeed, the Sindarin plural is formed this way precisely because an OS plural marker i is appended to the words, causing i-affection. However, for the present purpose it is also quite relevant that the verbal ending OS: -ia (derived from CE: -yâ) leads to i-affection without a plural. In fact, any ending adding an i would do, these are just the most common ones. However, there is a subtlety involved with the verbal ending:

In ordinary use, an OS verb does not end in the bare stem, but rather in an inflectional ending, e.g. OS: *baria-so (he protects). If we would do i-affection for the bare stem, the result would be **beiria, but in fact the a is not the last vowel (except for the i) if inflectional endings are present, and hence the i-affection yields beria-so (which subsequently undergoes further changes, see next section).

On the other hand, we can see the difference by considering OS: arani (kings). Here, i is really the last, one a is second to last, and therefore we produce an intermediate ereini (the change ei -> ai being a later development).

After i-affection has been carried out, any final vowel (combination) is lost in Sindarin, the only exception being vowels in monosyllabic words (even Sindarin doesn't produce words without vowels). Hence a pronoun ON: ho would not become h in Noldorin but remain N: ho (this might be S: so in Sindarin).

Thus, we move towards the plural formation pattern of Sindarin: In the above example, ereini would become erein.

Note that again a verbal stem (after i-affection) beria- would not really be found in this bare form in the language, hence the ending -ia is not lost but screened by an inflectional ending and therefore usually preserved (see below for details).

In a final step, ei becomes ai when in the final syllable, thus producing S: erain (kings). Note here the development of some seemingly irregular plurals: From OS: makla (sword), we find an intermediate meiklai. Loss of final vowels results in meikl, which is hard to pronounce, therefore in Sindarin an o is inserted. A final mutation then yields the plural S: meigol. This is most relevant for CE noun formation endings starting with -l or -r like CE: -la.

The diphthong ai is changed into ae.

2. Consonant development

Some of the consonant changes are simply changes in writing: In Sindarin, c rather than k in OS is written; final -v is written as -f instead, final -w (if preceded by a consonant) is written as -u. Several other consonant groups become slightly changes:

kkh -> ch
pph -> ph (f)
tth -> th
kh -> h

and initial s- is lost in the combinations

sth -> th (word initial)
sph -> ph (word initial)
skh -> ch (word initial)

The nasalized stops nd, mb and ñg are changed into stops in Sindarin, however they reappear when a word is mutated, giving rise to the special case class of mutations:

nd -> d
mb -> b
ñg -> g

For example, OS: ndair (bridegroom) becomes S: daer.

The most prominent change in consonants concerns single consonants and certain consonant groups following vowels: For single consonants, we see the effects of lention.

Hence, OS: beleka (mighty, great) becomes S: beleg (loss of final vowel and lenition k -> g), OS: *atan (man) becomes S: adan, OS: ndakro -> S: dagor

Longer consonant groups are more difficult. Some are evidently unchanged, even when preceded by a vowel:

mm -> mm
ss -> ss
nt -> nt
nd -> nd (sometimes nn)
ll -> ll
st -> st
nc -> nc
ng -> ng
ngw -> ngw

See e.g. OS: brasse -> S: brass (white heat)

Others experience lenition for the first consonant (if preceded by a vowel):

kl -> gl
kr -> gr
tl -> dl

For example, see OS: etled- -> S: edledhia- (go to exile), OS: ndakro -> S: dagor

For those involving liquids in trailing position, we see the effects of liquid mutation (as far as not yet carried out for p, t, k in the OS stage). Of course, the liquid sound is unaffected by a preceding vowel:

lm -> lv
rm -> rv
rg -> r
lg -> l
rn -> rn

See e.g. OS: parma -> S: parf (book), OS: dalma -> S: dalf (palm)

Others are assimilated

ld -> ll
mb -> mm
ks -> s (?)
dm -> nt

See e.g. OS: belda -> S: bell (strong) OS: lambe -> lamme -> lamm -> S: lam

One has to remember, however, that final -ss, -mm is often written as -s, -m, cf. OS: lambe -> lamm -> S: lam (tongue). A final -h is lost without trace. Furthermore, when not final, often the following assimilations occur:

nt -> nn
nd -> nn
nc -> ng

For example OS: *lambe -> *lamme -> *lamm -> S: lam (tongue), OS: *anta -> S: anna (to give).

As a side remark:This development presumably helps to shed some light omto the form S: lammen (my tongue) which has the Quenya equivalent Q: lambenya. Tentatively assuming the existence of some CE possessive ending -nya, -na relavant for Sindarin and Quenya (the precise form does not matter), we can explain the form as OS: *lambe-na -> *lamben -> S: lammen, i.e. the -e- appears as the preserved ending of CE: lambê rather than as part of some ending ?-en.

3. Final vowels in Sindarin

There are essentially three possibilities how a final vowel can be preserved in Sindarin:

a) in a monosyllabic word, cf. OS: *na -> S: na (to, towards)

b) as a final -w trailing a consonant, cf. OS: *tinwe -> tinw -> S: tinu (spark)

c) shielded by a -sV (where V represents any vowel).

The last combination is most relevant for verbs, because the 3rd person of a verb would involve an ending -so, -se, -sa. Hence, OS: linda-so -> linna-ho -> linna-h -> S: linna preserves the vowel. This is the reason that the final -a of A-verbs is in general not lost along with all the other final vowels.

4. Examples

Example 1: Starting from OS: *kaita, we change writing conventions k -> c, convert ai -> ae and do the lenition t -> d. Since we deal with a verb, we preserve the ending and find S: *caeda-.

Example 2: Writing ng -> g and doing i-affection o -> i yields OS: *ngolia- -> S: gelia- (learn). Again, we do not erase the ending. This verb is a special case mutation, hence we would find i ngelia (who learns).

Example 3: Writing nd -> d (and keeping in mind that this word will undergo special case mutation), we drop the final vowel, mutate b -> v and remember to write final -v as -f to find OS: *ndabe -> S: *daf (privilege).

Example 4: OS *meltha -> S: *meltha- (to seduce) remains unaltered.

Example 5: OS *rokla becomes, after loss of the final vowel and mutation rogl. As for OS: makla (sword), an o is inserted in Sindarin, resulting in S: *rogol (saddle). However, this word would not form its plural as **regyl but as *rygol.

Part IV: Acknowledgements

I am grateful to Helge Fauskanger for his essays on Primitive Elvish and Old Sindarin. In the preparation of this article I have made heavy use of the organized information in these works which made the use of the firsthand evidence from the Etymologies much easier.

Thorsten Renk


Back to the index