Mutations in Sindarin
PQ: Primitive Quendian
CE: Common Eldarin
SD: Sauron Defeated
PM: The Peoples of Middle-Earth
LR: The Lost Road
WJ: The War of the Jewels
SIL: The Silmarillion
LOTR: The Lord of the Rings
HOB: The Hobbit
RGEO: The Road goes ever on
VT: Vinyar Tengwar
PE: Parma Eldalamberon
Mutations are one of the most prominent features of Sindarin, yet at the same time one that is at times extremely difficult to understand. This is certainly in part due to the fact that mutations can arise for a number of quite different reasons. It is the aim of this article to present some facts and some plausible conclusions about the nature of mutations in Sindarin.
The origin of nearly all Sindarin mutations lies without a doubt in the sound changes in the development from primitive Elvish towards third-age Sindarin. In examples like
PQ: atar -> OS: atar -> S: adar (father) (LR:349)
CE: belekâ -> OS: beleka -> S: beleg (great) (LR:352)
CE: *kyelepê -> OS: kelepe -> S: celeb (silver) (LR:366)
we see sound changes for consonants following a vowel in the transition from OS to S which correspond to what we observe in mature Sindarin as lenition, in the above examples e.g.
t -> d
c -> g
p -> b
r -> r
l -> l
Similarly, the changes going from CE to OS reflect other types of mutations, see e.g. from
CE: alkwâ -> alpa -> OS: alpha -> S: alph (swan) (LR:348 )
lp -> lph
which corresponds to what is theorized in Sindarin to be liquid mutation or
CE: rattâ -> OS: rattha -> S: rath (course) (LR: 383)
tt -> th
for what is to become known as stop mutation in Sindarin.
The nasal mutation is more elusive: WJ:363 however reveals, discussing the class plural S: eledhrim (star-folk) that dhr < n-r in secondary contact. We may infer from this that it makes a difference if consonant clusters involving an -n- came into contact at the early CE stages or during compound formation in more recent developments. Indeed, the formation of the past tense is a good example for what happens to a cluster in 'first contact'. Here, we find S: pent (from the 'Túrin Wrapper') and Q: quentë (PM:401), the fact that nasal infixion is seen in both Sindarin and Quenya indicating that it reflects some CE form relevant for both languages. Here, we observe the cluster -nt which, according to the rules for nasal mutation should be changed into -th if it were relevant - but in fact, it isn't - since the cluster has been there since the CE stage, the rules of nasal mutation do not apply. The form govannen (and other examples) finally indicates that -nt can only be final in Sindarin, if it is moved into a word by an ending, it is changed into -nn-.
Since very few clear examples of compound objects in OS are known, not much can be inferred about the relevance of nasal mutation for the evolution of Elvish sounds from first-hand evidence. Among the few examples are the mentioned S: eledhrim and the name Caradhras (Redhorn).
We observe from these examples that liquid and stop mutation reflect the oldest changes from primitive Elvish to Old Sindarin, lenition and nasal mutation a more recent change in the transition from Old Sindarin to Sindarin, and finally the mixed mutation, having no clear counterpart in the outlined development must be the youngest type.
Part I - Phonological (P)-mutations
1. Prefixes ending in a vowel
We can immediately understand the relation of mutations in Sindarin to the sound shifts in the language evolution using the example of prefixes ending in a vowel. Take for example the word carth (deed). Using the negative prefix ú-, we create in an intermediate step *úcarth. However, now that the former initial c- has moved into the word and is in fact behind a vowel, we find structure that cannot be a Sindarin word that could have developed from OS, because this development would shift any consonant following a vowel according to the known pattern, here c -> g (see example above).
In order to meet the required structure of a Sindarin word, the -c- adjusts to the prefix and we find a mutated úgarth (VT44). We see the same development exemplified in ON: tre-batie -> S: trevedi (traverse) (LR:352) with the shift b -> v: there is little doubt that the literal meaning of this word is 'through-go' and that the un-prefixed element ON: batie would be S: bedi and not **vedi since no (unmutated) Sindarin verb or noun is known to begin with v-.
We see the prefix go- act in a similar way on the same element bad- in the greeting Mae govannen (well met) where we find the perfect passive participle go-vannen.
2. Analogical developments for prefixes ending in consonants
Does this imply that the compound of prefix and verb/noun has to exist in OS already in order to lead to mutation in Sindarin as the above example of S: trevedi might suggest? And is all we see in Sindarin then the regular sound development of the full compound? In other words, would a prefix that is created after the OS stage not lead to mutation?
Hardly. This is because we see S: ambenn (uphill) (LR:348 ) for which we can deduce from the Quenya forms that this has to be OS: *amba-penda. However, assuming that this is a compound in OS and carrying through the usual sound shifts would yield **ammabenn, not ambenn.
On the other hand, developing OS: *penda yields S: pend, penn and similarly carrying through sound shifts for OS: amba yields S: am. Therefore, although the prefix seems to be perceived as a separate element (it undergoes the sound shifts as if there were no word coming behind, making the loss of the final vowel possible), it still causes mutation in modern Sindarin.
Obviously many of the mutations are carried out as analogies - the prefixes act as if they still had a final vowel attached to them although there cannot be a continuous development from the ancient forms of prefix and noun/verb to the observed form.
Thus, every prefix that ended at the OS stage in a vowel could potentially cause lenition in mature Sindarin. This concerns dad- (down) (we do not know what the final vowel would have been, but dadbenn indicates that it is there), ab- (behind) from OS: *apa (attested in Quenya) and ath- (across) (from OS: *attha-).
3. Other mutation types for prefixes ending in consonants
Yet this is not the end of the story: SIL:428 has the prefix ar(a)- 'high, noble, royal'. According to what we said above, S: ar- could well represent a former OS: *ara- and we would expect the prefix to cause lenition. However, what we see instead is arphen (a noble) (WJ:376) from ar + pen with liquid mutation or Argonath from ar + gon(d) + ath without any mutation. Finally, there's the name Arwen where it is unclear if this is simply the ending -wen appended to the prefix or if this represents really OS: *ara-gwende (noble maiden) and we see lenition.
There is a second prefix ar- (beside, without) (LR:349). This leads to aronoded (without count) with liquid or soft mutation, arphent (and [it] said) with liquid mutation. However, here we do not have any reason to assume that there is an older form which ends in a vowel.
From those, we may deduce the following:
* There are prefixes which never ended in a vowel. Those presumably cause different types of mutation determined by their final consonant.
* Even prefixes which once ended in a vowel may or may not cause lenition - ar(a)- is clearly capable of causing liquid mutation, and the case might be similar for ad- (re-, back) which can be connected with Q: at(a)- (re-, back). This prefix might well be capable of causing stop mutation, very likely both lenition and stop mutation are possible. In principle, such arguments might apply also to S: ab- which reads in Quenya Q: apa- with variant Q: ep- - an indication for this is seen in WJ:387 where we see S: aphad(a)- from primitive ap-pata. Hence, something like 'after-word' could well be S: *apheth rather than S: ?abeth in Sindarin. Maybe the difference is that these are old compounds which have been preserved and not formations characteristic for Sindarin - this is very likely the case for aphada- -but we do not know in general, but it is hard to understand that S: arphen should be ancient where S: ambenn is not.
* Argonath, showing no mutation at all is clearly slightly odd in the context, but this is presumably a Gondorian word rather than a Sindarin one - the majority of prefixes does seem to cause any form of mutation.
4. The definite article
We can readily generalize the theory of mutations we have developed so far to include the definite article. As apparent from the tengwar version of the King's Letter in tehtar mode, the i-tehta of the definite article is not written on a separate carrier but directly above the first consonant if mutation is caused, see e.g. iVaranduiniant, iDrann.
This seems to indicate that the definite article acts in these cases just like any prefix - it attaches itself to the word in question, the former initial consonant of the word is now medial, this represents a structure which cannot be in Sindarin words developed from OS, hence a mutation in analogy to produce a word that could represent a development from OS occurs. This is also true in plural (the fact that the definite article is i in plural in Quenya but in in Sindarin clearly indicates that any consonant cluster created by compounding the definite plural article cannot be as old as the CE stage) where we see e.g. iMbair. Often, a hyphen or dot indicates something like compounding, e.g. Narn i-Chîn Hurin (Tale of the Children of Hurin), but we do not see this in all cases in the transcription, see e.g. Naur dan i ngaurhoth! (fire against the werewolves!), nevertheless we observe mutations in all cases.
This indicates that an object does not need to be a part of an actual compound in order to cause mutations - the possibility that the object could form a compound seems to be sufficient.
The fact that we observe objects like the definite article which seem to retain some degree of independence gives rise to the speculation that we may actually treat prepositions on the same footing as prefixes - as objects which are not forming actual compounds but could in principle. There are several points in favour of this view:
First, we have several cases in which the same element is attested both as preposition and as prefix, e.g. or-, or (above) (LR:379) or di-nguruthos (beneath death-shadow, prep.) (LOTR) and ar díheno ammen (and forgive us, prefix) (VT:44) or trî, tre- (through) (LR:392).
Second, in some cases we see the prepositions joined with a hyphen indicating some degree of compound formation, e.g. in di-nguruthos or na-chaered (into distance) (LOTR).
If so, we could expect that any preposition which ended in a vowel at the OS stage would cause lenition like any prefix that meets the same condition - in most cases, this is indeed what we observe even without hyphen, e.g. na vedui (at last) (LOTR). However, there are a number of exceptions, most of them in a single text, the 'Ae Adar' (VT:44). We will discuss them carefully after discussing the regular pattern and give arguments why many prepositions in this text do not seem to follow the pattern indicated in other sources.
This implies essentially the same for prepositions whose mutation pattern is unknown as for prefixes - the mutation might be determined by the final consonant or, by analoguous mutation, the preposition might cause simply lenition. Probably, one should choose whatever sounds better (I have the strong suspicion that is precisely what Tolkien did). As an example, athan (beyond) (SD:62) is probably capable of causing both nasal mutation and lenition. If we knew that the stem did never end with a vowel, we could reduce this probably to nasal mutation only.
Some words of warning: Although we treat prepositions for the purpose of mutations as a kind of 'loose prefix', this does not imply that any prefix can be used as a preposition or that prepositions and prefixes are always the same.
In WJ:367, Tolkien explicitly remarks that primitive WO did not yield an independent word in Quenya but only the prefix ó- (o-). In the subsequent paragraph dealing with the Sindarin evolution of this element, go- and gwa- are several times referred to as prefixes, strongly suggesting that there is no independent word **go in Sindarin either. Sometimes, we observe that an independent element takes a meaning different from the prefix: In Quenya, we find Q: úquétima (impossible to speak) (WJ:370) but Q: ú (without) (VT:39) as independent word.
Especially considering the 'Ae adar', you are not required to follow the main line of reasoning here that prepositions are for mutation purpose nothing but prefixes in invisible compounds. However, if you do not accept this, you should throw away your mutation chart immediately: About 80% of the entries (including the whole liquid mutation and all but a few examples of the stop mutation) are based on precisely this conjecture - if you reject it, please get an empty sheet of paper and start collecting what we really know of independent prepositions causing mutations. It is my view, however, that the general structure of the argument holds and that what we see in the 'Ae Adar' can mostly be explained by carefully examining every case.
6. Conditions for compound formation
In spite of the fact that prepositions and the definite article show some degree of independence from their nouns, they presumably cause mutations as if they were inside compounds. This, however, is intuitively clear when one considers that preposition + noun or article + noun would be perceived as the the first step in assembling a complete sentence from individual words, a combination 'the tree' or 'above clouds' somehow belongs together, whereas 'the new' or 'abvoe now' in some sense does not belong together. Intuitively, we would expect for the idea of invisible compound formation to make sense that no mutation would occur when a word intrudes between definite article and noun or preposition and noun.
We do not have an example of an adjective separating definite article and noun, but we know what happens in several cases when an article intrudes between preposition and noun - this leads to compound formation of preposition and article, cf. nu + in -> nuin (under the) (SIL), or + i -> erin (on the) (SD:128 ), o + i -> uin (from the) (SD:128 ), all these objects causing mixed mutation. This implies exactly what the name indicates - the force of the preposition is somehow broken up by the intruding object, but since it formes a compound, the resulting mutation is sometimes that of the preposition, sometimes that of the article.
An example of a different kind is provided by Naur dan i ngaurhoth! (Fire against the werewolf-host!). This example suggests along with aglar 'ni Pheriannath that prepositions ending in -n and causing nasal mutation do not append an extra -(i)n (and do not form compounds) since that is their natural ending already. Here we see that the mutation is not determined by the preposition any more but by the intruding article (it would be i gaurhoth otherwise). Note that the singular article is used here to denote the collective plural in -hoth. This is apparent from Tol-in-Gaurhoth (isle of the werewolves) (SIL) where we see the plural article causing nasal mutation. We may conclude from this example that a preposition can only cause mutation for a word following immediately behind, despite its logical relation to the noun. If anything separates preposition and noun, any mutation of the noun is no longer determined by the preposition. We may similarly conclude that anything separating article and noun (like an adjective coming between) would make a compound formation impossible and no mutation could be caused by the article on the noun.
The interesting question is: would article and preposition cause mutations on the intruding object instead? In other words, are they objects 'eager to mutate anything on their right hand side' or are they picky and require a close connection with the following word? Clearly, 'the new' does not have the same closeness as 'the man'.
7. The relative pronoun
We cannot answer this question from the examples investigated so far, however we note that the relative pronoun i(n) seems to share the properties of the definite article: the natural partner of the article is the noun, however the natural partner of the relative pronoun is the verb, 'who sees' is the minimal version of a relative sentence and therefore if any object is closely associated with the relative pronoun, it would have to be the verb. Indeed, we observe nasal mutation in Dor Gyrth i chuinar (Land of the Dead that live) (Letters:417) or lenition in Dor Firn i Guinar (Land of the Dead that live) (SIL) or, employing what is presumably a compound of an + i lenition (mixed mutation?) in ai gerir (for those who do) (VT:44). (Note that there are again dissenting examples in the 'Ae Adar' which we will discuss later on).
Assuming that in the standard scenario a relative pronoun acts on a verb by the same rules as an article acts on a noun, we do have one example where the relative pronoun meets something different from a verb: i sennui Panthael estathar aen (lit. presumbaly 'whom they preferably should name Fullwise') (SD:128 ). This would be clearly in line with the assumption that the relative pronoun does not simply lenit 'anything' but is a very selective object waiting for a verb.
Generalizing this example, this would imply *Dor Gyrth i si cuinar (Land of the dead who live now) or *i beren benn (the bold man) with the adjective in an extremely unusual position (see further down). However, usually it is very hard to separate articles and prepositions from their nouns just because they are so closely related that they tend to form compounds. What we have discussed here is possibly most relevant for relative pronouns where it is easiest to slip in an adverb.
8. Preposition as adverbs
An immediate consequence of this is that if prepositions are used as adverbs, i.e. an (for) or dan (but, however) as used by several writers based on Quenya examples, they would lack the logical partner (the noun) they have in their role as prepositions. In consequence, they could not form compounds and, since apparently these elements do not lenit anything regardless of word type, no mutations would be caused.
There is a neat example in English illustrating the difference: English 'for now' can have two distinct meanings: 1) 'for the moment' 2) 'because now'. In the first case, we see an inseparable preposition-noun compound, one cannot move the 'now' in the sentence 'for now I've had enough' without altering the meaning; 'for I had enough now' is not the same. Therefore, this might be *as si in Sindarin (nasal mutation).
On the other hand, 'for (because) now he has come' is prefectly well with moving 'now' to the rear: 'for he has come now' means precisely the same thing. Therefore, this represents two adverbs and would be *an si in Sindarin (no mutation).
9. Compound formations for nouns
In discussing how prefixes and elements similar to prefixes end up when connected to verbs and nouns, we found the following: If the first object ends with a vowel, lenition is always carried out. If the first object ends with a consonant, other mutation types may occur, however often lenition is carried out (we observed even different behaviour for the same prefix!) in analogy to other elements, loosely spoken under the general assumption that most elements would have had a final vowel in the past and without tracing what the true evolution history of a compound would have resulted in.
In dealing with compounds of two nouns or noun and adjective, we may expect exactly the same behaviour. However, there are hardly any nouns in Sindarin ending with a vowel, so in most cases we would expect several examples to exhibit lenition and others to show other mutation types.
In discussing compound formation, we do not make any real difference in the following between compounds where the words are really written together, such as Mordor, or where the compound nature is indicated by a hyphen, such as Gil-Galad. We do this because we have seen in the case of articles and prepositions that hyphenation is apparently used to indicate compound formation. Most of the compounds are names, and the formation of names is a very complex issue with Tolkien sometimes changing the explanation for a name after some time. Often, Quenya elements intrude in names and are blended with Sindarin. We will not attempt to discuss all possible compounds but just give several examples to show the range of possible outcomes.
A fair share of compounds exhibits mutation types other than lenition. Here, we find mor + dor -> Mordor (Darkland) (SIL) with liquid mutation (the first element is explicitly listed as mor and morannon (black gate) indicates that this is indeed mor and not morn), similarly mor + caint -> morchaint (darkshapes) (SIL), one explanation of Barad-dûr (Dark tower) would see this as a compound with stop mutation (for a different explanation see adjectives further down), in Caradhras (Redhorn) (LOTR) we see nasal mutation, likewise in eledhrim (star-folk) (WJ:363).
However, there is also a fair share of compounds showing lenition by analogy, in spite of the fact that the final consonant would require different mutation, among them Gil-Galad ('starlight', at least according to UT:65. In PM:347 however, the name is explained as 'star of radiance' introducing a new word galad, one of the mentioned mind-changes of Tolkien regarding names), calen + sad -> calenhad instead of **calessad (greenspace) (UT:425), nin + talf -> Nindalf instead of Nithalf (Wetwang) (A Tolkien Compass: 195) or gil + taur -> Gildor instead of **Gilthor (Starlord) or celebrin + tal -> Celebrindal (Silver-foot) (SIL).
We may think of establishing the time period of a compound formation by studying its mutation pattern. Consider for example Angband (Iron Prison) (SIL) ,Angmar (Iron Home) (LOTR) and Anghabar (Iron-delvings) (SIL), all with the same initial element ang- from CE: angâ (PM:347) and yet in the first case, no mutation is caused whereas the second two lenition is the result.
Evidently, Angband is a very old structure (it is for sure older than the elves...). Therefore, the compound word probably stems from the CE stage already and would probably be CE: angambandô. This directly developed into S: Angband, preserving the b- in analogy with CE: andambundâ -> S: annabon (elephant) (LR:372).
On the other hand, Anghabar (persumably derived from a root SAPA 'dig, excavate' (PE12:82)), is a structure of Gondolin at a time where Sindarin was completely developed already - therefore lenition by analogy would have been in place (and there would never have been a CE form **angasapâ or such like), resulting in Anghabar instead of **Angsabar. Similar arguments hold for the even later kingdom Angmar which is a modern compound, lenited by analogy, and not a descendant of a CE compound C**angambarê.
We may map out later changes in the difference between Calenhad and Caradhras. Colors in PQ are usually derived using a suffix -i, this is changed into -e in CE and lost in OS, hence we may assume that both prefixed elements had no final vowel in OS already, hence OS: calen and OS: caran. However, an OS compound OS: calensad would be subject to the rules of nasal mutation in compound words and end up in S: **calessad - hence the compound has to me more recent than the OS development phase. On the other hand, Caradhras does exhibit nasal mutation and may therefore be a true OS compound or may be a later compound not subject to lenition by analogy.
There are a few compounds in Sindarin which do not seem to show any mutation, among them the river names Anduin, Baranduin, Esgalduin and Taur-im-Duinath (SIL). However, these could potentially all result in lenition of d -> dh which Tolkien though of as 'uncouth' (UT:267) (if you try to pronouce some of them carrying out the lenition you might get an idea why...). As we will see below, he frequently did not carry out lenition for this particular letter.
10. Summary phonological mutations
We have seen that a large class of mutations seen in Sindarin one common explanation can be given - all of these mutations are caused because compounding (visible or invisible) occurs. All these mutations share some characteristical traits:
* they are caused in combination of two words
* the last letter of the first word determines the type of mutation of the second word
* if this letter is a vowel, lenition occurs
* if this letter is a consonant, lenition may still occur by analogy, however other types of mutation are then possible and occur frequently
* if a prefixed object did never end with a vowel, presumably lenition cannot occur
We are not able to predict reliably if a given element ending with a consonant would cause lenition or other mutation for the following word - this seems to be dependent on both the period in which the compound was formed and on the second element. Most compounds seem to reflect a complex development history, but lacking background information we cannot in general trace the evolution of compounds reliably. Presumably, it's best to choose the most pleasant sounding combination in hand-made compounds.
In most cases, at least any kind of mutation is caused, for some of the exceptions explanations can be given. It seems unlikely that a particular preposition would generally not cause any mutation.
Part II - Grammatical (G)-mutations
So far, we have encountered mutations as regular sound-shifts for compounds from an early stage of Sindarin and mutations carried out by analogy to those. However, considerting the sentence Daur a Berhael, Conin en Annûn, eglerio! (Frodo and Sam, princes of the west, glorify (them)!) (LOTR) we see something different which is commonly interpreted as lenition of direct objects.
For the moment, let us that the first name represents a lenited form of *Taur (noble one). The underlying form is unattested and we will discuss this assumption in the next section. But if this name is lenited, we see a mutation that can quite evidently not be caused by a word forming the second part of a compound, because the word in question is at the beginning of a sentence and there is no word before. Hence this belongs to a group of mutations which is quite different to all we have seen before.
Presumably, the lenition of *Taur is caused by the grammatical function of the name as object - objects seem to be lenited to make them distinct from the subject of the sentence. In the following, we will call this a grammatical mutation, because unlike the phonological mutations it cannot be linked to the phonological development of the language.
Since grammatical mutations cannot be linked to a specific word triggering them (by acting as the first part if a compound), there is no final consonant or vowel which might determine the type of the mutation - consequently we cannot have nasal, stop, mixed or liquid mutation in a G-mutation - only one type may ever occur: lenition.
We may therefore imagine the following development of the mutations leading further and further from their origin in the sound shifts in the evolution of Sindarin:
Mutations carried out as regular sound shifts for compounds formed in CE or OS (in strict sense, those are not mutations in Sindarin)
Mutations carried out in analogy for compound formed in Sindarin where the final consonant determines mutation (P-mutation)
Mutations carried out in analogy for compounds formed in Sindarin where lenition is assumed by default (P-mutation)
Mutations carried out to signify a grammatical function (G-mutation)
The latter can be motivated by the fact that in a more abstract sense, the object joins the verb to form a sentence not quite unlike the prepositions joins a noun to form an object - only that compounds on the level of sentence structure are much more loose and less well-defined entities.
2. Lenition of direct object vs. alternatives
We may observe lenition of direct objects in the following sentences (not including those where lenition would be without effect):
Lasto beth lammen! (Hear the word of my tongue!) (LOTR)
Im Narvi hain echant (I Narvi made them.) (LOTR)
Daur a Berhael, Conin en Annûn, eglerio! (Frodo and Sam, princes of the west, glorify (them)!) (LOTR)
Ar e aníra ennas suilannad mhellyn în phain (And he desires to greet there all his friends.) (SD:128 ) (mh- is a variant of v- for lenition of m)
caro den i innas lin (make it thy will) (VT44)
No lenition is carried out for the following, although they might be seen as objects:
si loth a galadh lasto dîn! (untranslated, probably 'now shall flower and tree listen to silence')
Lacho calad! Drego morn! (Flame Light! Flee Night!) (UT:65)
pedo mellon a minno (Speak 'friend' and enter!) (LOTR)
edregol e aníra tírad i Cherdir Perhael (...) Condir i Drann (In especial he desires to see Master Samwise (...), Mayor of the Shire) (SD:128 )
Ar e aníra ennas suilannad mhellyn în phain (And he desires to greet there all his friends.) (SD:128 ).
Daur a Berhael, Conin en Annûn, eglerio! (Frodo and Sam, princes of the west, glorify (them)!) (LOTR)
Let us investigate the case for lenition of direct objects by trying to find other explanations for the attested six examples: hain does not necessarily seem to imply a pronoun as direct object - LR:385 contains N: hein which would correpsong to S: hain and indicates that in fact all Noldorin pronouns of this kind begin with h- in spite of the fact that they are derived from the stem S-. Similarly, the underlying form of den is unattested, and although a form in t- is very likely, this is not proof.
The underlying name *Taur is not attested separately as well - if this where ?Daur to start with, only lenition of Perhael (which is attested separately in SD:128 ) is to be explained. If we try the compound theory, then the conjunction a (and) would cause lenition for the following object. This, however, is not seen in pedo mellon a minno. We might still get by using the conjecture that a causes lenition only for nouns. However, in loth a galadh we see again contradicting evidence. Evidently, such a theory does not fare too well and is better discarded.
Lasto beth and caro den could be explained by assuming the imperative compounds with the following word and causes lenition. This would also be in agreement with Edro hi ammen! (Open now for us!) (LOTR). It is contradicted by pedo mellon (speak 'friend'), but as we will discuss below, the Moria gate riddle seems to imply a confusion between a quotation mode (unlenited) and a vocative (unlenited) - in any case the sequence 'm-e-l-l-o-n' probably has to appear as such on the gate for the riddle to make any sense. Furthermore, we will see in the discussion of adjectives that lenition of m- is often ignored, so this may not mean much. Imperative lenition is also contradicted by lasto dîn, but this may not mean much either since this also contradicts lenition of direct objects - is is probably the 'uncouth' dh- again which is avoided here. But a truly serious blow for imperative lenition by compound formation is Lacho calad! Drego morn! where we cannot dismiss the first example so easily. Since lenition following imperatives cannot account for this example and is for sure not relavant for Daur a Berhael (the imperative would have to come before the words to form a compound) we may finally dismiss it as well.
Lenition of the verb object is the only theory that can satisfactorily account for all examples where we see lenition of nouns which are not part of a compound.
Let us next examine the examples which contradict this hypothesis. We have already discussed lasto dîn. From the next example Lacho calad! Drego morn!, we may deduce that Sindarin has a vocative which remains unlenited, hence 'Flame o light! Flee o night!'. This, in fact, is nicely in line with the Moria Gate riddle which Gandalf had read as vocative 'Speak, o friend, and enter!' first but which in truth was a quotation mode 'Speak 'friend' and enter.' which also would have to be unlenited.
The next two examples aníra suilannad and aníra tírad concern gerunds rather than nouns. We have already argued in the case of P-mutations for the relative pronoun that mutations are not carried out for any word but expect a certain type of word. If grammatical mutations are of similar origin, they can be expected to share this property, and these examples are most easily explained by taking the gerunds as infinitives here rather than as nouns and assuming that lenition of objects is carried out for nouns only, hence 'wishes to see' rather than 'wishes the seeing'. (However, sometimes gerunds appear in a role very close to nouns, cf. Mereth Aderthad (feast of reuniting) (SIL) where a translation as infinitive makes little sense. We have to assume that gerunds in this function would have to be lenited like ordinary nouns.)
Finally, from the last example we can conclude something about the grammar of object lenition: Only the first (or the first few) in a row of objects is lenited. This would be quite in line with the fact that in Quenya in a row of objects inflected for a certain case only the last word in the row gets a case inflection ending, cf. Q: Namna Finwë Míriello (the statute of Finwë and Míriel) where the genitive -o is only appended to the last word.
3. The rules of direct object lenition
Accpeting lenition of verb objects as the most plausible explanation of what we see in the corpus, what are its rules?
We have already established that for a longer object, only the first word is lenited. This shows clearly in edregol e aníra tírad i Cherdir Perhael where we see unlenited Perhael (SD:128 ). However, we see in Daur a Berhael, Conin en Annûn, eglerio! that the conjunction a (and) is apparently able to start a new object which is lenited again. Since we have established with some certainty that we indeed observe a grammatical mutation here, we can assume that this is independent of a as such and that the same would also be true for ar (and) and egor (or), hence *Daur egor Berhael...eglerio! should also be acceptable. And yet we see that if a longer phrase intrudes, no 'new' object is perceived and consequently no lenition is carried out after a conjunction, even when a new object stands behind it: edregol e aníra tírad i Cherdir Perhael ... ar Meril bess dîn ... ar Baravorn (In especial he desires to see Master Samwise ... and Rose, his wife...and Hamfast) (SD:128 ). Apparently, the lenition of direct objects has a rather short memory and 'forgets' that nouns are still objects of the verb if too many words intrude. In practice, lenition of an object after a conjunction is probably optional.
In the majority of cases, dative in Sindarin is expressed using an (for), e.g. Anno ammen sír i mbas ilaurui vín (Give us today our daily bread.) (VT:44). It is however permissible to express dative without this marker just like a direct object as Ónen i-Estel Edain (LOTR) shows. As the two valid translations 'I gave the men hope.' and 'I gave hope to the men.' indicate, this is so in English as well, word order marking the difference between dative and accusative (not however that in English dative comes before accusative if expressed without preposition whereas in Sindarin the accusative comes first). As Ar e aníra ennas suilannad mhellyn... shows, dative objects expressed without an are subject to lenition just as accusative objects: suilannad is transparently a compound suil + anna- (give greeting) and the verb 'to give' has the person who receives something in dative by the very definition of dative (latin 'dare' to give, hence 'dative' = 'give-case').
However, in reverse we may infer from A Pherhael ar am Meril suilad (To Sam and Rose [be] greeting) (SD:128 ) that if any other object is formed using a preposition or case marker like an that no lenition of objects is carried out - what we see here instead is the P-mutation of the preposition forming a compound with the names. We see in teithant i thiw hin (drew these signs) (LOTR) that the same is true for the definite article. Generalizing, we may assume that P-mutations always have stronger force than G-mutations: Whenever an object would be subject to a P-mutation leading to some different result than lenition, the P-mutation is carried out.
We may infer from the general line of reasoning that an object cannot be mutated twice, this would violate the idea of mutation as being something that is carried out in analogy to sound changes in the evolution of Sindarin. Hence, 'I eat the soup.' would be Medin i halph. and not **Medin i chalph., the definite article inducing mutation i halph and the grammatical lenition of the direct object being lost. For the case marker an, this is evident e.g. from am Meril (not **am Veril).
We have already discussed the fact that the lenition of objects does not touch verbforms such as the gerund used as infinitive. The example egor ben genediad Drannail (or in the Shire-reckoning) (SD:128 ) shows that a preposition as initial element of an object is not lenited (we know from Quenya the preposition Q: ve (as, like) which is the phonetic equivalent of be. This can be interpreted in two ways: We may assume that the grammatical mutation of an object is only carried out for direct objects (those not expressed with a preposition) or that prepositions (like verb forms) are not among the objects subject to this type of mutation. The latter assumption fits better with the general set of ideas developed here because the joining of an indirect object to the rest of the sentence is conceptually to really different from the joining of a direct object and because we know already that a G-mutation is not carried out for every word type. As it is, however, we are unable to decide this question now (and the distinction has little importance for writing Sindarin sentences which is fine with just the fact that there's no additional lenition for indirect objects without explanation).
4. An investigation of adjectives
Common lore has it that adjectives follow a noun and are lenited in this position. Yet a thorough investigation of the actual data tells a slightly different story:
We find 10 examples where we cannot decide if lenition is carried out:
annon edhellen (elvish gate) (LOTR)
Amon Rûdh (bald hill) (SIL)
Amon Ereb (lonely hill) (SIL)
Dagor Aglareb (glorious battle) (SIL)
Ered Engrin (iron mountains) (SIL)
Ered Luin (blue mountain) (SIL)
Iant Iaur (old bridge) (SIL)
Noegyth Nibin (Petty Dwarves) (SIL)
Bair annui (Western Lands) (SD:128 )
o galadhremmin ennorath (from tree-woven Middle-earth)(LOTR)
The last exampleis included here because the framework above would interpret this as unlenited since the preposition cannot cause mutation for anything but a noun, however the explanation that this is stop mutation carried out on galadhremmin cannot be proven wrong. In any case, it is the only case in which an adjective precedes the noun and we may take it as a hint (but not more) that in this position adjectives are not lenited.
In 9 examples lenition is carried out:
Ard-galen (green realm) (SIL)
pinnath gelin (green ridges) (SIL)
Eryn Vorn (dark forest) (LOTR)
Curunir 'lan (white wizard) (LOTR)
Tol Galen (green island) (SIL)
Ered 'Wethrin (shadowy mountains) (SIL)
Talath Dirnen (guarded plain) (SIL)
Lô Dhaer (Great Fen) (VT42:15)
genediad drannail (shire-type reckoning) (SD:128 )
In 9 cases, no lenition is carried out:
Ered Mithren (grey mountains) (LOTR)
Cú Beleg (great bow) (SIL)
Talath Rhunen (eastern vale) (SIL)
Imloth Melui (lovely flower-vale) (LOTR)
Rath Dínen (silent street) (LOTR)
Dor Dínen (silent land) (SIL)
Barad-Dûr (dark tower) (LOTR)
Côf gwaeren Bêl (the windy Bay of Bêl) (VT42:15)
Nan Tathren (willowy vale) (SIL)
In addition there are the two more complicated cases involving adjective and possessive adjective
i mbas ilaurui vín (our daily bread) (VT44)
mellyn în phain (all his friends) (SD:128 ).
where the first might hint at lenition. The second example is a clear nasal mutation. We will deal with those after we have understood the behaviour of a single adjective trailing a noun.
Apparently the vast majority of adjectives comes indeed behind a noun, but judging from the raw data, only about half of them are lenited, the other half is not. This is not too strong a case for lenition of adjectives. Both the list of leniting adjectives and of non-leniting adjectives contains some dubious examples: tirnen might be seen as a participle rather than an adjective, *trannail is based on the conjecture that i drann is indeed i + trann and not i + drann with failed lenition of d -> dh. Ard-galen might be seen as a compound rather than an adjective. Regarding the list of lenition failures, it is somewhat unclear what rh- would lenit into. The example rhass -> i-rass (LR:363) suggests that this should become Talath Runen, Helge Fauskanger argues in 'Sindarin - the Noble Tongue' rather for rh -> thr but in any case we should see something. In WJ:333 we find Dor Dhínen instead of Dor Dínen where the lenition is carried out. Finally, we already discussed that Barad-Dûr can well be understood as a compound rather than an adjective trailing a noun.
However, as the examples Dor Dhínen and Dor Dínen suggest, we may possible clean the list somewhat by removing these examples where we consistently find that lenition is not carried out. It has further been suggested that the lenition m -> v is sometimes ignored as well - this is possible, however even if this is so (and the data as such provide little basis for this speculation since in P-mutations we usually see this being carried out), it is applied in a less consistent way than in the case of d - dh and we are still left with three unexplained examples indicating that the lenition of adjectives is not a rock-solid rule but apparently to some degree optional (unless one wants to include a set of rules that it is also permissible to ignore b ->v, rh -> r (thr) and t -> d). Considering this, probably there's nothing too special about m -> v since we have examples of t -> d (un)applied as well, but one can never know. In the following, we will continue under the main assumption that usually m -> v is carried out but also investigate the consequences of this being optional. Cleaning both lists of suspicious candidates, we end up with a 6:4 advantage for lenition, in other words, 60% of all adjectives seem to be lenited. This cannot be called a very reliable rule.
5. Are adjectives P or G mutations?
Considering examples like calenhad and our uncertainty in the identification of Barad-Dûr and Ard-Galen, the immediate question might be if adjectives are not in fact broken compounds with the opposite word order, i.e. Eryn-vorn.
This view is hard to reject outright (and it is probably partially correct). However, what leaps to the mind is the following difference between compounds and the adjectives: In discussing compounds, we found that a sizeable portion of objects (about one third to one half) was showing mutations different from lenition, e.g. Caradhras (Redhorn) with nasal mutation. If adjectives were of the same origin, we'd expect to find a similar share of different mutations. What we find is, however, just one among 28 examples mellyn în phain with anything different from lenition and this is among the complicated examples (and might have a different explanation to be discussed later). There are several clear cases which could exhibit something different from lenition but don't, among them Tol Galen (not **Tol Chalen), Eryn Vorn (not **Eryn Morn) and Ered 'Wethrin (not Ered Gwethrin).
Thus, lenition by analogy occurs in a far larger proportion than for compounds which indicates that adjectives are at least somewhere between broken compounds and true grammatical mutations. For the following rules of adjective lenition, we are unfortunately dependent on how far we believe the mutation of adjectives to be a G-mutation already.
6. Possessive and demonstrative adjectives
In order to proceed, we first have to establish the role of the possessives: We find examples with nouns determined by possessives alone in the corpus:
Meril bess dîn (Rose his wife) (SD:128 )
sellath dîn (all his daughters) (SD:128 )
ionnath dîn (all his sons) (SD:128 )
ered e-mbar nín (the mountains of my home) (UT:40)
Adar nín (my father) (VT44)
i eneth lín (thy will) (VT:44)
i arnad lín (thy kingdom) (VT:44)
i úgerth vin (our sins) (VT:44)
i innas lin (thy will) (VT:44)
Could the last two examples rather be lín and vín? Probably the difference is not easy to spot in Tolkien's handwriting. From rather general considerations which somewhat exceed the scope of this article but are covered in CE views on the Sindarin pronomial system we can assume that dîn represents a lenited form of *tîn and vîn a lenited form of mîn (the latter is actually attested in VT:44).
In addition there is the demonstrative which we can assume to be lenited from sin (this is Q: sina in Quenya):
i thiw hin (these signs) (LOTR)
Since these objects behave in all respect like we'd expect from adjectives, i.e. they follow the noun and are lenited in this position (indeed, they seem to be lenited even more reliably than adjectives) it is probably reasonable to treat them as such. Therefore, we will assume in the following that for questions of mutations we can treat possessives like adjectives.
7. The rules of adjective mutations
We have already established that adjectives usually follow the noun and stand a fair chance of being lenited in this position, however this is not mandatory and can probably be chosen according to what yields the most pleasant sound.
We have argued that the fact that only lenition seems to occur for adjectives trailing a noun (and indeed the fact that often no mutation at all occurs) favour an interpretation of adjectives as true grammatical mutations instead of one as broken compounds. Turning to the two examples involving two adjectives, we find the only real counterexample to this view:
i mbas ilaurui vín (our daily bread) (VT44)
mellyn în phain (all his friends) (SD:128 ).
Clearly, phain has not undergone lenition but nasal mutation. It is possible that this is a speciality of the possessives, but a simpler explanation for both examples is to visualize them as a two step compounding process:
First, all adjectives are joined together, undergoing P-mutations. Hence, for the fictional example 'all brave men' we might compound beren + pain -> beren phain. This compound is then (like a single adjective) joined to the noun as a G-mutation, hence *binn veren phain (and not, as a second P-mutation would cause, **binn meren phain). This makes some sense, as it helps to resolve the clear mismatch between an apparent nasal mutation and the good arguments for the interpretation of adjective lenition as a G-mutation.
What happens if we break the sequence by a conjunction, i.e. 'a brave and good man'? Clearly, no P-mutation among the adjectives is possible any more (we have seem above that P-mutations cannot be carried out across intruding words), but the second adjective might be subject to optional G-mutation again, much as the verb object is after a conjunction. Hence we might assume that benn veren ar vaer and benn veren ar maer are both acceptable (and so would benn beren ar maer if we do not carry out lenition, also I would not use it like this).
We might also have an example of what happens if the adjective is far from its noun, but this requires a leap of faith (and might have an entirely different explanation, see further below). This is because it involves a participle rather than an adjective:
A Elbereth Gilthoniel o menel palan-diriel (Oh Elbereth Starkindler from heaven gazing-afar) (LOTR)
Technically, palan-diriel is used as an attribute of Elbereth, but it is unlenited. This may indicate that lenition is lost if too many words intrude between noun and adjective (even for a G-mutation), or that lenition of adjectives is, after all, still relative close to P-mutations or that palan-diriel remains unlenited for an entirely different reason. As it is, we cannot decide the question on the basis of this rather uncertain example.
Above, we argued that a preposition should not form compounds with adjectives and hence should not cause mutation. If this is true, o galadhremmin ennorath (from tree-woven middle-earth) (LOTR) shows an adjective that could still be subject to a G-mutation (adjective describing noun), but in fact isn't. This tells us that the adjective coming before the noun might be unlenited (and this agrees nicely with the evidence from adverbs further below). Adjectives in this position are, however, extremely unusual as the number of examples indicates.
8. The adjective in predicative use
We have one example of a Noldorin sentence showing an adjective in predicative use:
Lheben teil brann i annon (untranslated, probably 'five feet high [is] the door') (Tolkien - Artist and Illustrator)
Apart from the interesting fact that the verb 'to be' is absent in Sindarin, this shows an unlenited adjective brann. This might be due to several reasons, from this one example we cannot tell:
* predicative adjectives might be lenited as the attributive adjectives discussed above, but since the adjective in this particular example is before the noun, it remains unlenited
* predicative adjectives might be unlenited in general
* predicative adjectives might always be in front position to make a distinction and therefore always be unlenited
A common assumption is that predicative adjectives are in general unlenited, since there is a gap created by the absence of the word 'to be' (which must have been present in primitive Elvish) and it is intuitively plausible that a mutation could not develop over this gap. This finds a nice analogy in the lenition in Irish:
Rinne sé an scian géar (He made the knife sharp = He sharpened the knife; predicative use, unlenited)
Rinne sé an scian ghéar (He made the sharp knife; attributive use, lenited)
While this is a plausible scenario for Sindarin also, no definitive conclusions can be drawn at present.
9. Lenition of adverbs
Adjectives cannot only be used to describe a noun but also to describe a verb as adverbs in Sindarin. By analogy, we could expect that adjectives following a verb and describing it as adverb are lenited, adjectives coming before a verb and describing it as adverbs might remain unlenited and so might adjectives used as adverbs if they are too far from the verb. By further analogy, we can hope to extend these rules also to participles and adverbs like 'now' which cannot be used as adjectives. Let's take a look at the corpus to test these expectations:
We find an example of an adverb preceding the verb, unlenited:
Mae govannen! (Well met!) (LOTR)
We also find several adverbs following the verb. For some of them, no statement can be made:
Noro lim! (untranslated, probably 'run on', 'run fast') (LOTR)
Cuio i Pheriain anann! (may the Halflings live long) (LOTR)
Ar e aníra ennas (And he desires there) (LOTR)
Some show lenition when following the verb:
edro hi (open now) (LOTR)
síla díriel (untranslated, probably 'shines watchful') (Lays of Beleriand 354)
Some do not:
le nallon sí (to thee I cry now) (LOTR)
Anno ammen sír (give us today) (VT:44)
le linnathon nef aear, sí nef aearon (to thee I will sing, on this side of the Sea, here on this side of the Ocean)(LOTR)
Finally, there is a complex example with two adverbs (one an adjective, one a participle, both unlenited)
silivren penna míriel (silver glittering slants down jewel-like-sparkling) (LOTR)
We do not have as much examples as for adjectives, but let us nevertheless try to clean the lists from all dubious examples to determine what percentage of the adverbs is lenited. In the case of adjectives, we removed those starting with m-, this eliminates ...penna míriel. We might remove participles all together, then we are left with edro hi, le nallon sí, Anno ammen sír and le linnathon nef aear, sí nef aearon. But we'd almost expect that the last two examples do not show lenition since words intrude between verb and adverb. The case thus stands 1:1, i.e. half of the adverbs following the verb is lenited, and this is about what we see for adjectives. Including the participles *tíriel and míriel leaves the ratio where it is.
Apart from the fact that we have less statistics, adverbs do not seem to be drastically different from adjectives: Most of them come behind their verb, they show about the same ratio of lenition when they follow the verb, they apparently are unlenited when they come before the verb (this is about the only difference - adverbs can apparently easier moved into front position), they remain unlenited when too many words intrude between verb and adverb and the lenition of m- may or may not be carried out.
It has been suggested that only adverbs following an imperative are lenited and that the lenition of tíriel has a different reason. While this is hard to counter (after all, the theory is exactly tailored to what we observe), it fails to pass Ockham's razor which states that from two explanations, the less complicated one is to be preferred. Since the data suggest that adverbs and adjectives can be treated on the same grounds, there's not need to assume a special role for the imperative (and we have already rejected a general lenition following imperatives).
10. Lenition of verbs
We have already discussed the lenition of verbs as P-mutation caused by the relative pronoun. The corpus contains one example that might hint at lenition of verbs as G-mutation:
Guren bêd enni (My heart tells me.) (VT:41)
It has been suggested that verbs directly following the subject of the sentence are lenited. Apparently, we do not see nasal mutation here, so this might be a G-mutation. Since in Celebrimbor o Eregion teithant the verb is unlenited, the rule seems to be similar to what we have seem above again: If words intrude between subject and verb, the mutation is lost.
This simply picture is, however, explicitly contradicted by the inofficial
A Rian pent (And Rian said)
where no lenition of the verb is observed, in spite of the fact that it follows the subject. At this point, we severely suffer from lack of any evidence (and in fact, even analogies to guide us). There remain essentially two possibilities:
* the lenition of pêd might be caused by the possessive ending -n (my) which is related to Quenya -nya and hence can be expected to have ended with a vowel in the past
* the lenition of verbs following a noun might be optional
We do not know enough to make a decision here.
11. Summary grammatical mutations
We have argued that there is a large second class of mutations that is not caused by any specific word. Since no final consonant can be responsible for mutations in this case, we always see lenition. The prototype of this G-mutation is the lenition of verb objects. We have inverted the argument to determine of the mutations seen for adjectives and adverbs are G- or P-type mutations and we found that they can better be described as G-mutations, although they are apparently still close to broken compounds in some respect.
G-mutations can be seen as analogies to the formation of compounds on the level of whole sentences. Often, they do not seem to be mandatory, and we encounter unlenited words more often than in the case of P-mutations. Their detailed rules are often complex.
For nouns, G-mutations seem to occur relatively predictable. For adjectives, we could draw relatively firm conclusions due to the large number of examples. Here, lenition seems to be the rule, but to some degree optional. For adverbs, we lack the number of examples to draw similarly firm conclusions, we could however show that there's no indication that they are substantially different from adjectives (apart from the fact that adjectives describe a noun whereas adverbs describe a verb).
Finally, no conclusion could be reached regarding the behaviour of verbs.
Part III - Exceptions and irregularities
1. Is the 'Ae Adar' special?
The system of mutations outlined so far is not free of exceptions, indeed, we found several occasions in which words were not lenited in spite of a rule indicating mutation. Mostly, this was found for the more loose grammatical mutations, there are very few examples (except the case of d -> dh) where mutation was completely ignored.
There is, however, one text which contains several examples in striking disagreement with the theory developed so far. This is the Sindarin version of the Lord's prayer, the 'Ae Adar', published in VT44. It is the observation that all severe exceptions to the outlined scenario occur in one draft text (the translation of the prayer is not finished) that has given rise to the position of this article to treat the mutation pattern of the 'Ae Adar' as irregular - a plausible scenario is that Tolkien simply tested some alternative ideas.
A different position has been brought forward by Aaron Shaw (private communication, yet unpublished in closed form), treating the 'Ae Adar' as regular development and deducing the rules of mutation based on this assumption. While this approach is hard to prove wrong, I reject it for the following reasons:
The set of rules governing mutations has to be described in the framework of modern syntax theory, an approach that does not show up in Tolkien's own discussions of his languages and was possibly not known to him. Using this framework, several phrases can be explained but would have extremely awkward literal translations, something that is not encountered in other, better understood Sindarin and Quenya texts. The ultimate set of rules is quite complex and close to the point where one rule is needed to describe one occurence of mutation, i.e. the theory becomes purely descriptive and lacks predictive power. Some of the rules (like lenition as a marker of abnormal word order) allow easy construction of highly ambiguous sentences if no further rules to prevent this are introduced. Nevertheless, it is an interesting alternative approach and the reader is encouraged to make up his own mind.
In the following, we will examine the 'Ae Adar' carefully, point out the irregularities and try to find plausible explanations based on what we have deduced so far.
2. Indeclinable forms
Already the first phrase of the prayer shows an extremely puzzling construction: Ae adar i vi Menel (our father who [is] in heaven). This phrase does not just violate one but three of our expectations:
1) i as the relative pronoun should cause mutation for verbs only, even more so since there's a gap created by a lost form of 'to be' between i and the preposition.
2) A preposition should not be subject to lenition at all (we can deduce that this is indeed mi from its Quenya counterpart mi, the fact that no other unlenited word in Sindarin is known to begin with v- and the attested name Minhiriath) 3).
3) A preposition ending with a vowel should cause mutation for Menel.
We essentially see the same problem in the later phrase bo Ceven sui vi Menel (on earth like in heaven). Here:
1) bo should, as preposition ending with a vowel, cause mutation for Ceven
2) sui should not be able to mutate another preposition (indeed, it fails to lenit mín in a later phrase)
3) vi should cause mutation for Menel again.
The mere fact that not only one of the rules deduced above is violated by the forms but several makes it suggestive to assume that there's indeed something special about these forms which causes a chain of alterations of the rules.
Both Menel and Ceven are capitalized, indicating that they are seen as names rather than as nouns. The entry KEM in the Etymologies (written much earlier) might provide a clue to the behaviour of these forms:
KEM- soil, earth Q kén (kemen). N coe earth (indeclinable), cef soil, pl. ceif (...)
In a language like Noldorin/Sindarin without case inflection and for a form like coe which cannot form a distinct plural in any case, 'indeclinable' can only have one more possible interpretation: not subject to mutation. This means that Tolkien very likely had some nouns in mind which could not be mutated, and one of them is derived from the same stem from which the later Ceven is derived (Ceven itself is not as a Noldorin form in the Etymologies).. This makes the comclusion rather plausible that Tolkien actually kept this idea in his later works but transferred it to the new word Ceven he had derived for Sindarin. The fact that in the prayer both Ceven and Menel appear unmutated and that they signify conceptually similar things (separate 'realms' of the creation) makes it plausible to apply the same explanation to Menel where we do not have similar clues.
Hence we might have solved one problem - but we are still left with the expectation that we'd find ?mi Menel instead of vi Menel. Seemingly, the relative pronoun and sui cause mutations on the prepositions - but do they indeed? If we take sui mín as example, sui does not necessarily seem to cause mutation. One might argue that the lenition m -> v is not carried out in all cases anyway, but it is still hard to understand that sui + mi should be lenited whereas sui + mín should not.
Assuming for a moment that it is not sui that causes lenition on mi, it must be lenited for quite a different reason peculiar to the phrase vi Menel as such, and hence the relative pronoun would also follow our expectation and not cause lenition. We know nothing of the grammar of the conjectured indeclinable forms - it may well be that something has to be lenited - so the lenition would 'jump' to the preposition, hence not **mi Venel (because it's indeclinable) and not **mi Menel (because the phrase is after all object and something has to be lenited) but vi Menel (according to my subjective impression, this happens to be the most pleasant-sounding combination of the three possibilities). If this is so, we might assume that bo Ceven likewise represents *po Ceven - we don't have this preposition attested otherwise (Quenya has an element Q: po, opo (after) as a variant of Q: apa- though).
3. Alternatives to indeclinables
An alternative approach would see the preposition bo arising from older *bog - the final -g being lost in the language evolution but conceiveably creating a preposition which would not cause mutation. While this would work for bo Ceven (and is of course hard to dismiss in any case), it cannot explain vi Menel where we happen to know the underlying form. We might assume that Menel here just represents an instance of an unapplied m -> v lenition, but we have to explain the lenition of vi then. The only choice left is then to assume that the relative pronoun and other prepositions can indeed lenit words other than verbs and nouns respectively, but this immediately causes the need to find an explanation for i sennui in SD:128 and i gohenam (who forgive) to be discussed further down.
One might simply assume that bo, vi are unlenited prepositions which for some or the other reason also do not cause mutations. However, we have already discussed the rather strong evidence that vi is indeed a lenited form of mi.
4. The question of sui
We see sui (like) act two times on different objects: sui vi Menel (like in heaven) and sui mín (like we). Based on the Quenya versions of the Lord's Prayer (the Ataremma in VT43) which in all versions shows an emphatic pronoun Q: emme (we) where Sindarin has mín I have argued in CE views on the Sindarin pronomial system that this is an emphatic nominative form 'ourselves' in Sindarin also. If so, it would be important to make it distinct from the possessive adjective mín (our) (the latter object is attested in the 'Ae Adar' also). Hence, there's actually good reasom not to carry out a normal lenition for sui. The alternative would be that sui is derived in such a way that it does not carry out lenition. It might be related to what appears in the Ataremma at this position, i.e. Q: sivë (as, like), but this rather seems to be sin + be with abnormal sound shifts.
As it is, we cannot descide the question and are faced with two not too charming alternatives: sui might not cause lenition on a general basis (which would be possible but hard to argue) or it might cause lention in general but this is ignored to mark nominative case here. Probably, the latter explanation is simpler and hence to be preferred.
5. Lenition of prefixes
Immediately following, we see the phrase i gohenam di ai gerir (who forgive those who do). Evidently, no lenition is carried out for gohenam, neither is nasal mutation if the relative pronoun would be plural. One could dismiss this by arguing that Tolkien just had no lenition after relative pronouns in mind when he wrote the text, but this is not true - ai gerir (for those who do) apparently does show lenition after the form ai which is presumably a compound an + i(n) (for those who) involving that dative marker. This shows that there has to be a specific reason why the form is not lenited.
We note further that the verb 'to forgive' appears earlier in the text as díhena-, whereas all versions of the 'Ataremma' have the same verb avatyar- or apsen- in both positions. In the Sindarin version however, we find di immediately following after gohena-. It has been argued that this is the pronoun 'those' referring to the sins. If so, it would be used as an object here, however, and we see in the same text the form den (it), indicating that the object pronouns end with the regular ending -n in this text as well (see CE views on the Sindarin pronomial system for details). Could this di the former prefix, replaced by a new one instead?
We know form contrasting ortheri (LR:395) vs. esgeri (LR:379) that prefixes can be subject to i-affection or not. We have argued above that it is possible that prepositions are not subject to G-mutations. If so, we may extend this to hold for prefixes as well - as long as they are perceived as prefixes and not as part of the word like in esgeri. This may be one explanation for the lack of lenition of palan-díriel discussed above - but this is circumstancial evidence at best.
Assuming that there is indeed a class of prefixes which is not seen as part of a word (and not subject to i-affection), adding yet another prefix to such a word poses some difficulty. Assume we'd want to use go + díhenam (with the intended meaning 'we all together forgive' ). We cannot have **godhihenam (because the prefix doesn't lenit), **diohenam (we can assume that since the prefix is seen as prefix, the order or prefixes plays no role) sounds somewhat strange, so in fact 'postfixing the former prefix might provide a viable alternative, gohenam di does not sound too bad.
There is no conclusive proof for this, all that can be shown that if a prefix is subject to i-affection (e.g. go. in genediad (reckoning)) then it is also subject to mutation (e.g. go- in aronoded (uncounted)). It is, however, a possible explanation for the strange lack of lenition in i gohenam and at the same time for the occurance of di.
We do see a prefix lenited in Iarwain ben-adar (oldest without father), Tom Bombadil's Elvish name though. But it is most likely that this is not so much supposed to represent the preposition 'without' as rather the adjective ending '-less' (why else would it be lenited), as such, it is probably rather close to the noun in spite of the hyphen and should rather be read *penadar. Well - this is not as one would wish it, but hardly any explanation fares better in the face of these irregularities.
6. Some alternatives
Clearly the picture outlined is somewhat unsatisfactorily because many of the interpretations appear rather introduced ad hoc.
What are the alternatives if one tries to accept the mutation seen in the 'Ae Adar' as regular? Trying to reconcile the evidence of i chuinar with i gohenam, one might arrive that the conclusion that in addition to i as article and i as relative pronoun, there's a third such object not causing mutation for the following word. How can we predict its appearance? Well, in the two cases where we do not see mutation, i sennui Panthael estathar and i gohenam, there is no need to interpret the relative pronoun as subject of the relative sentence (in the first case because it is presumably object 'whom [they] preferably should call Panthael' and in the second case because the subject is in the verbal ending -m). There's no definite bit of evidence for or against this interpretation, so it is in fact a viable alternative. However, an additional explanation of the form di in the 'Ae Adar' (or if it is taken to be a pronoun, for the lack of the ending -n) has to be provided. Accepting this explanation, we furthermore have to do away with the idea that mutations only concern certain types of words because of i vi Menel (the relative pronoun here being subject and according to this interpretation causing mutation for any following word). This, however, does not agree well with the apparent lack of lenition seen in aníra suilannad and aníra tirad, so a special rule making these cases different is required.
One can explain the lack of mutation seen in sui mín by assuming that this represents nasal mutation sui i mín -> suin mín -> sui mín carried out on the possessive. This, however, creates the somewhat awkward literal translations 'like the ours' which is, I have been told, acceptable in New York slang, but this is not something I would expect Tolkien to build his languages from.
In general, things do not appear to get nicer if one tests other theories. Partially, we can understand what we see in the 'Ae Adar' by the assumption of indeclinable objects. Partially, it is probably best to assume that Tolkien tested different ideas of what causes mutation in this text which do not in all cases agree well with what he had used in other texts.
Part IV - Acknowledgements
I would like to thank Lothenon and Eirien for early discussions of their and my own ideas about mutations in all shades and Aaron Shaw for introducing me to modern syntax theory (which I found very inspiring although I do not believe in the immediate relevance for Sindarin).
Back to the index