As I mentioned in one of the comments on the previous blog post, one of the characteristics I invented for neo-Khuzdul is a prefix marking the definite accusative — that is, it comes before a noun which is the direct object of a verb, if that noun is not being newly introduced into the universe of discourse — that is, it has been previously mentioned or implied, or can be assumed to be well-known to the person or persons spoken to. Discussing this is moving a bit ahead and outside of chronological sequence, but as this is a detail which doesn’t have major effects on the rest of the language, it seems to do no harm to treat it out of order.

I have received a couple of questions about this prefix. The Dwarrow Scholar asks:

 What is the neo-Khuzdul prefix in question please (or is it indeed “al” just as in Arabic) ?

The prefix is (in theory) id-, and appears as such before a word beginning with a vowel (e.g. id-urus “the fire”) but it appears assimilated to following stops, e.g. ib-bekâr “the weapons.” Whether it assimilates to other types of consonants I’m not yet sure; probably it does assimilate to nasals (m, n) but not to liquids or glides (l, r, y).

This prefix certainly resembles the Hebrew prefix ʔeṯ– and I can hardly doubt that it was inspired by it, as Mad Latinist suggests; yet what I actually remember from the time when I invented it was thinking of the Persian definite accusative suffix –rā. Most likely the fresher memory of the one and a somewhat dimmer memory of the other combined to influence this choice.

Helge asks:

Is the variation in form of the definite article/prefix meant to reflect actual phonological changes (like say, assimilation) taking place over time? If I had been developing Khuzdul, I would be careful not to presuppose any really substantial “developments”, since Tolkien insists that this was a language that largely resisted change. (The only attested “sound-change” is that the preposition aya can be reduced to ai!) I like the idea that F may represent original P (a nod to Arabic), but to suggest that there was an older stage where consonantal roots were connected with a distinct “characteristic vowel” (as in Adunaic) hints at a pretty substantial structural change taking place over the course of history. Isn’t this more dramatic change than what Tolkien seems to presuppose? Do you have an vision of what the originial “Pure Aulean” Proto-Khuzdul was like as well as your suggestion for the “historical” version?

I have not explored the internal history of neo-Khuzdul to any great extent, and I’ve assumed that its current form is not unlike that of its original form. Accordingly, I used Arabic rather than Hebrew or Aramaic as an inspiration, because Arabic, at least in its classical form, is very archaic and conservative in structure. But there are two reasons to suppose that Khuzdul ought not to be constructed as if it had never undergone any change. The first is that Aulë, as a language-creator was completely capable of building in elements that resemble the processes of language change, even if they had never taken place in history, and that, if he was anything like Tolkien, he probably did! The second is that although change in Khuzdul was slow and slight, it was not nonexistent: “After their awakening this language (as all languages and all other things in Arda) changed in time, and divergently in the mansions that were far sundered… the change in Khuzdul… was ‘like the weathering of hard rock compared with the melting of snow.'”

Accordingly, an assimilation here or there hardly seems like an outrageous development to postulate. In Arabic, although for the most part root-consonants remain intact without assimilation, some affixes do assimilate; notably the definite prefix al-, which assimilates to following coronal consonants, and also the infixed –t– of the derived verb stem conventionally numbered VIII, which assimilates in voice and emphasis to a preceding coronal obstruent. Whether the Khuzdul assimilation took place in Longbeard Dwarvish over the long years between the awakening of Durin and the end of the Third Age, or whether it was something which Aulë/Mahal built into the language from the beginning is a question I haven’t felt the need to answer definitively.

The first major text which I had to translate into neo-Khuzdul, and the last really significant work I did on the language prior to The Hobbit was Durin’s Song/The Balrog, the lyric which plays in the the background of the Moria scene in The Fellowship of the Ring. I received the request for this on February 27, 2001 (although, due to time zone issues, it was dated February 28), and returned a reply on March 4 — an unusually long delay of a little over 4 days, of which doubtless not all was spent working on this text.

The original English text, written by Philippa Boyens, was as follows:

Durin who is Deathless / Eldest of all Fathers. / Who awoke / To darkness / Beneath the mountain / Who walked alone / Through halls of stone

Durin who is Deathless / Lord of Khazad-dum / Who cleaved / The Dark / And broke / The silence/  This is your light! / This is your word!/ This is your glory! /  The Dwarrowdelf of Khazad-dum!

A crown of stars in the cold, black water of Kheled-zaram. Durin sleeps.

Deeper into the earth/ The dark grows heavy/ Cold snaps our bones/  Deeper into the earth. / There, the glint of Mithrail / sharp and far away/  Deeper into the earth/ That sound again / Dread surrounds us. /  Can no one hear us?A great shadow / Moves in the dark/ The earth shakes! / Cracks! Splits! / Will no one save us?! / Fire! / Fire in the deep! / Flames lick our skin! / Fear rips our heart! / No! No! No! /  The demon comes!

My translation was:

Durin ku bin-amrad / Ugmal sullu addad / Ku bakana / Ana aznân / Undu abad / Ku ganaga / Tur ganâd abanul

Durin ku bin-amrad / Uzbad Khazaddûmu / Ku baraka / Aznân / ra karaka / atkât / ala lukhudizu! / ala galabizu! / ala ukratizu! / Khazad-dûm!

Kilmîn thatur ni zâram kalil ra narag, Kheled-zâram. Durin tazlifi.

Ubzar ni kâmin / Aznân taburrudi / Iklal tanzifi bashukimâ / Ubzar ni kâmin / gilim Sanzigil / shakar ra udlag / Ubzar ni kâmin / tada aklat gagin / Ugrûd tashurrukimâ / Maku kataklutimâ? / Askad gabil / Tashfati ni aznân / Kâmin takalladi / Tabriki! Takarraki! / Maku zatansasimâ? / Urus! / Urus ni buzra! / Arrâs talbabi fillumâ! / Ugrûd tashniki kurdumâ! / Lu! Lu! Lu! / Urkhas tanakhi!

This was a pretty literal translation, with a few necessary reductions of redundancies, e.g. “the Dwarrowdelf of Khazad-dûm.”

Analysing the features of this text will take some space, yet as it was foundational for the later versions of neo-Khuzdul, it deserves to be treated in depth.

Let’s first look at the verbs, since we have already covered their forms to some degree. They are:
bakana “awoke”
ganaga “walked”
baraka “cleaved”
karaka “broke”
tazlifi “sleeps”
taburrudi “grows heavy”
tanzifi “snaps”
tashurrukimâ “surrounds us”
kataklutimâ “can hear us”
tashfati “moves”
takalladi “shakes”
tabriki “cracks”
takarraki “splits”
zatansasimâ “will save us”
talbabi “lick”
tashniki “rips”
tanakhi “comes”

All of these are 3rd persons, mostly singular. They fall into two obvious classes: “perfects” like bakana, ganaga, baraka, karaka, mostly translating preterites, but only because these refer to historical facts. The others, “imperfects” containing the prefix ta– refer to present experiences or present or future possibilities. It will be observed that three of the forms contain the 1st person plural suffix –mâ, which doubles as a possessive marker “our” with nouns and as an object suffix “us” with verbs.

The verb roots are heavily laden with “jests,” most of which should be obvious to those familiar with the history of Germanic languages, or even just with English. However, I had better go over them, since the puns may be somewhat less obvious in the present constructions. For the most part I can remember the sources easily.

BKN “wake” is from Gothic gawaknan “awaken” and of course English waken, with substitution of B for W, which doesn’t exist in neo-Khuzdul.
BRK “cleave” from Gothic brikan, English break.
GNG “walk” from Gothic gangan “go.”
KLT “hear” from Indo-European *klutos “heard.”
KRK “break” from English crack.
NSS “save” from Gothic nasjan “save.”
ShFT “move” from English shift.
ZLF “sleep” from Gothic slêpan “sleep”, whose preterite is saizlêp.

Please note that these are not intended to suggest any historical or other relationship between Khuzdul and these languages — they were simply sound-sequences that seemed appropriate at the time.

Others are from Tolkien languages:
BRD “grow heavy” from Adunaic burôda “heavy.”
LBB “lick” from Eldarin LAB “lick” (though this is also an Indo-European root of the same meaning).
NKh “come” from Adûnaic unakkha “he came.”

Others appear to be pure inventions, or at least I cannot remember the source or association with certainty. Perhaps a perceptive reader can figure them out!
KLD “shake”
NZF “snap” — possibly simply the consonants of “snap” rearranged and altered.
ShNK “rip” — this might be onomatopœic, from a sound of tearing, shnik!
ShRK “surround” — possibly from a badly maltreated Latin circum “around.”

Four of the examples show a doubled medial consonant: taburrudi “grows heavy,” tashurrukimâ “surrounds us,” takalladi “shakes,” and takarraki “splits.” This was supposed to be an auxiliary stem indicating long-continued, repeating, or otherwise extreme action: e.g. takalladi “shakes over and over,” takarraki “splits into many small pieces, ‘shivers’.”

Also of interest are the prefixes ka– and za-. These mean, respectively, “can” and “will/shall”, and their forms were suggested by can and shall – or perhaps, in the latter case, German sollen. Their usage is very un-Semitic, and for that matter rather un-Indo-European. I may have imagined them as reduced auxiliary verbs that eventually got attached to verbs as clitics; as they refer to potential or future states, which are certainly non-factual, they are attached to the “imperfect” verbs.

Helge wrote:

Our one-and-only Tolkienian Khuzdul pronoun is mênu, accusative “you”, from the battle-cry meaning “the Dwarves are upon you”. Your 2nd person inflections contain nothing readily similar. Then again, the Hebrew 2nd person inflections (like –ta in the 2nd person sg. perfect) are not similar to the independent pronoun for accusative “you” (ending –kha added to an accusative particle).

This is true, and it was never my intention to have the verb forms exactly mirror the independent pronouns or pronominal affixes (about which more will be explained it is place). However, you bring up something else interesting and problematic for neo-Khuzdul.

The phrase Khazâd ai-mênu “The Dwarves are upon you” has been well-known for nearly sixty years. Yet for most of that time one could only conjecture how ai-mênu meant “upon you.” It was quite possible, for instance, that ai was “you” and mênu was a postposition. Or perhaps ai was “they are” and mênu was an inflected form of “you”. This state of ignorance still prevailed when I started creating neo-Khuzdul. Therefore I simply disregarded these words, fearing more to mischaracterize them than to create a system which omitted them.

By the time we found out that ai was a clipping or combining form of aya “upon” and that mênu was the accusative of a plural “you” (Parma Eldalamberon #17, p. 85), I had already established a detailed pronominal morphology for neo-Khuzdul in which the independent form of “you plural” is astun (feminine astin). In all likelihood (I do not remember the details) –st– was a strengthening of 2nd person –s-, while –u– and –i– were masculine and feminine elements, and –n was obviously a plural element — the 2nd person singulars are astu/asti.

I don’t apologize for making forms inconsistent with ai-mênu — for the reasons I mentioned, it seemed more prudent at the time. My mistake was that when I learned about the meaning of mênu, I did not at once go back and try to find a way to fit it in with the morphology. At the time, however, the three Lord of the Rings films had been produced, there was no prospect of any more films, and I shelved neo-Khuzdul without expecting to do any more work on it ever. By the time I had to start work on neo-Khuzdul again, I was concentrating on making it consistent (insofar as possible) with the earlier work, and I neglected to note that there had been an inconsistency which I could have fixed. As a result, I created several phrases containing neo-Khuzdul 2nd person pronoun forms which are consistent with my earlier pronominal morphology, but not with mênu.

This was unquestionably an error on my part, a serious oversight — the more so because it concerns the most famous phrase in the Dwarf-language! It is not, however, an irreparable error. In fact, it creates an opportunity to expand and enrich neo-Khuzdul’s pronoun system.

Tolkien in multiple places indicates that both Elvish and Mannish languages possessed a distinction between two types of 2nd person pronoun: one formal/respectful/courteous/polite/deferential, the other familiar/imperious/endearing. I do not recall Tolkien saying anything about Khuzdul having such a distinction, but he also never says that Khuzdul doesn’t; and it provides a neat way of getting out of my self-inflicted 2nd person trap. The distinction need not have been an original Khuzdul one; it might, perhaps, been imitated from other languages, using an appropriate noun or title to fill out one of the 2nd person slots, much like Spanish usted and Portuguese você (< vuestra merced/vossa mercê), or the Quenya use of the ending –tar “high one, lord” (in some paradigms) to create honorific verbal forms.

Coincidentally helpful is the fact that mênu fits with certain established facts about neo-Khuzdul. The –u ending can be taken as the same as that seen in Khazad-dûmu — an accusative ending following verbs and verblike forms. In fact, it is quite possible that aya is really a verbal root “go over, be above, be superior to.” That leaves mên, of which the –n ending is the same as the existing pronominal plural ending in neo-Khuzdul.

The question now is to which category to assign mên — formal or familiar? There are valid arguments for both. The you-pronoun in Khazâd ai-mênu refers to hated enemies such as Orcs. If the familiar form is one exclusively used for endearments or for close personal friends, then presumably the formal would have to used for Orcs, whom one presumes the Dwarves would not tutoyer, as they say in French: to treat someone as such an intimate that one uses familiar pronouns with them. On the other hand, if the distinction is not one of familiarity vs. unfamiliarity, but of respect vs. the absence of respect (if not disrespect), then presumably the Orcs would get hailed with the less respectful pronoun.

Something very like this has happened in the history of English. The Old English pronouns þū (>thou) and gē (>ye) simply distinguished singular (one “you”) from plural (many “yous”). In the later Middle Ages, however — probably through imitation of French — thou was used for intimates, ye (accusative you) in formal situations for singulars as well as plurals. One used thou to speak to sweethearts, children, animals — and to God. But thou was also used for enemies, as a sort of insult, as if to suggest that one’s foe was no better than a child. In Thomas Malory’s Le Morte Darthur, in the Tale of Sir Gareth (which I use because it is probably not a translation from French) Gareth always respectfully addresses the damnsel Lyonet who accompanies him as ye, whereas she (assuming him to be a kitchen boy) calls him thou; when Gareth fights with other knights, they address each other as thou; but when Gareth’s enemies yield and offer him homage, their relationship is changed, and they now call each other ye. By the 17th century in standard English thou had fallen out of ordinary use, and the accusative you was replacing ye in all situations; thou only remained in archaic, especially religious language, and in some non-standard dialects (both geographic and class-based), where the actual form in use was a little different, e.g. tha or thee.

Since I also have to keep consistency (if possible) with my established usage, I have just checked to see where I used astun and related forms (e.g. the pronominal suffix –zun). It looks like it was primarily in situations that can be described as military, where one dwarf is ordering or encouraging another to perform some action. These cases would seem to fit the “respectful” profile. Therefore I conclude that astun is the respectful 2nd person plural, and mên is the familiar (if not disrespectful!) 2nd person plural, probably with mê as a singular form. I did, unfortunately already have a word mê “we” already in the pronominal paradigm, but since it doesn’t appear to have been used anywhere, that doesn’t create any particular problem; I’ll just have to create a new pronoun in its place, perhaps ammâ.

This is all new — I hadn’t really thought about the issue until last week, when Helge’s question forced me to consider the discrepancy. But the error has serendipitously enriched neo-Khuzdul, making it both more complex (and therefore more natural) and more consistent with Tolkien’s Khuzdul.

Helge writes:

I believe you once mentioned that you had started to work out verb paradigms for Movie Khuzdul? One has to imagine a Hebrew- or Arabic-like system, with varous “conjugations” (a “qal” of simple verb, a corresponding causative, the passive equivalents of both, and possibly even intensive conjugations).

The concept, and to an extent the forms of the Khuzdul verb, as I worked it out some thirteen years ago, were considerably influenced by the structure of the Semitic verbs, particularly Arabic.

Semitic verbs, like other parts of the language, are generally based on triliteral roots. From each root a number of bases can be formed, which allow for verbal formations like passives, causatives, iteratives, reflexives, and so forth. In Arabic there are ten normal ways in which such bases can be formed, in Hebrew seven, in Aramaic usually six (though in more ancient forms of Aramaic there were more). Not all of these bases are exemplified for each root, and their meanings are not always predictable. Although they are generally grouped together in dictionaries, to a certain extent they can act like independent verbs, not necessarily more closely related than such English verbs as conceive, deceive, receive, and perceive. Sometimes the basic form of the verb (which in Hebrew is called the qal form) doesn’t even exist, just as there’s no such verb as **ceive in English.

Within each base, there are forms which carry some of the qualities which in Indo-European languages are allocated to tense, mood, and aspect. The main distinction in most Semitic languages is between perfect forms and imperfect forms. Arabic has several other forms which — from the point of view of their shape alone — can be considered as variations of the imperfect.

Defining the difference between perfect and imperfect is a task of extreme complexity. The uses are different in the different Semitic languages, and they have also changed over time. One might say that perfect refers to actions which are over and done, while imperfect refers to actions that are in the process of happening, or are going to happen, but that would be a drastic oversimplification and in many respects would be inaccurate. It is, in any case, not really relevant to Khuzdul since, although I postulated a distinction that was formally similar to the perfect-imperfect distinction in Semitic, it ended up being functionally different.

However, the formal parallels are relevant, and to demonstrate them I’ll give an example of the perfect forms of the Arabic simple stem of the root KTB “write”:

Perfect

Other than the 2nd and 3rd persons having a distinction between masculine and feminine subjects, this actually looks a lot like an Indo-European verb. There’s a basic stem katab-, and all of the information about person and number is provided by suffixes. (Arabic also has dual verb forms, but I haven’t shown them because I never created dual forms for Khuzdul.)

If we turn to the imperfect forms of the same verb we see something quite different from what we’re accustomed to see in Indo-European languages:

Imperfect

Here we see that the base form is different, ktub instead of katab, and that the job of distinguishing person and number forms is borne not just by suffixes but also prefixes. In some cases we can attribute some separate meaning to each affix; for instance, ya– is 3rd person masculine, but ta– doubles in function as both 3rd person feminine and 2nd person general, while there is no common prefix for the 1st person forms. There’s a consistent suffix set marking masculine plurals (-ûna) and feminine plurals (-na), but all of the other forms end in –u, except for the 2nd feminine singular. It is rather a messy system, and kind of hard to memorize.

All of this was in the front of my mind when I started designing the Khuzdul verbal system. Let’s take a look at the Khuzdul forms comparable to the “perfect,” using the root ZRB “write, inscribe.”

The superficial similarities are obviously very close. For starters, I imitated the Semitic characteristic of having distinctively feminine 2nd person and 3rd person forms. Unfortunately (perhaps), most of these forms never got used: the 2nd person forms because there aren’t any female Dwarf characters (we think!) for a Dwarf to speak Khuzdul to; and the 3rd person forms because I never worked out which nouns would be feminine. In fact, I think that as with most languages of Middle-earth, masculine and feminine are not lexical properties; the only grammatical gender is “natural gender,” which could distinguish a male person or animal from a female, but not otherwise. I suppose the so-called “masculine” in Khuzdul is really a masculine/neuter, or a default form, while the “feminine” (if it really exists) is the marked form; a state of affairs which is objectionable to my ideas of social fairness and structural balance, but which is probably to be expected in a society where males outnumber females by two to one.
The base patterns correspond exactly to Semitic, being CaCaC. The suffixes are pretty self-explanatory; the endings –mi and –mâ suggest a 1st person element –m-. The second person is marked by an element –s-, and then –u and –i mark masculine and feminine. Second and third person plurals are marked by –n.
Zarabôn and zarabên must be *zaraba-un, *zaraba-in, with contraction of the diphthongs *au, *ai > ô, ê. The 3rd feminine ending –ai is distinctive, and is probably not from *zaraba-i (which would have given zarabê) but *zaraba-ai.

These forms, in meaning, are not comparable to the Semitic perfect. The Khuzdul “perfect” is not a past tense, nor does it necessarily refer to completed action. Rather, it refers to actions which can be considered as dependable facts, as opposed to evolving and uncertain realities. These might be statements about the past, such as one might find in a chronicle, or statements of general truth, such as Izgil taraza zann ra zann: “The Moon rises every night” (literally “night and night”, sc. one night after another) or Uslukh sharaga “A dragon lies” — i.e., comtinually, compulsively, and dependably. It’s the sort of form that would be used in an aphorism. It could also be used to describe events that will predictably and with certainty take place in the future: Durin zabakana “Durin will awake” — to the Dwarves, an undoubted fact about the future.

The Khuzdul forms corresponding to the imperfect of the root ZRB are as follows:

In a sense, this can be looked on as a partial rationalization of the Semitic imperfect. The person/number forms are still defined by a combination of suffixes and prefixes, but there is a consistent pattern: sa– marks 2nd persons, ta– 3rd persons; the suffixes are predictably -i, -în, -iya, -iyan. Only the 1st person plural breaks the pattern, and that because a 1st person plural is not, strictly speaking, a plural of the 1st person singular, but a 1st+2nd or 1st+3rd form. It will be noted that s– in a prefix in the “imperfect” corresponds to an –s– in a suffix in the “perfect.”

The stem, as in Semitic, is CCVC — in this root, ZRB, the stem vowel happens to be –a– (-zrab-) but in other roots it could be different. For instance, “I am writing” is azrabi, but “I am sleeping”, from the root ZLF, is azlifi. This is a purely lexical distinction, is unpredictable, and does not correspond to any kind of semantic class. It may point to a period in the past in which (as in Eldarin and Adûnaic) vowel distinctions were an integral part of the root; however, other than in these forms, no trace of this remains in Khuzdul.

The meaning of the Khuzdul “imperfect” is also different from its Semitic counterpart. It refers, not to incomplete action, but to vividly imagined action — either because one sees it directly in front of one, or imagines it as something which is playing out in the mind’s eye. It has no regard to tense. A Dwarvish storyteller would use this form to describe events he wanted his audience to vicariously experience, regardless of whether they had happened in the distant past or were prophecies of the future. It can also be used to describe an ongoing action that is taking place at the present: Durin tazlifi “Durin is (now) sleeping.” Durin zalafa could mean “Durin typically sleeps, as a matter of course” and would be a rather insulting thing to say to a dwarf; though in the right context, it could mean “Durin slept,” as an historical fact.

All of this was, of course, only the beginning; as I developed the verb, more and more complications arose, and the newly-invented forms often do not look anything like Arabic or any Semitic language.

Having already considered the methods I would use to extract meaning from various Khuzdul words, let’s go ahead and look at some of the words that seemed most easy to interpret. This is most easily done in a table:

But there remained a lot of words whose patterns couldn’t be readily interpreted, or whose patterns overlapped in an unclear way: CiCiC was used, for instance, for sigin “long”, but also zigil “silver” (referring to a silver-like color) and kibil “silver,” the actual metal. CaCaC appeared in baraz “red” and narag “black”, suggesting an adjectival formation — but also in zahar “excavation” or “underground dwelling”.

Prefixes and suffixes weren’t terribly clear either. Ma- I was pretty sure was a participial suffix, considering its similarity to Arabic mu-. A lengthened vowel + n was common in placenames: Nargûn “Mordor” next to narag “black”; Gabilân “Great River”, the Gelion; and Nulukkhizdîn, which I guessed meant “Dwarf-home on the Narog.” But then there was also Tharkûn “staff-man.”

I was pretty much on my own from that point on. I had a basic idea of how Khuzdul was shaped in my head, but specifics were elusive. The only grammatical facts available were that there were genitival or adjectival formations ending in -ul and that compound words were common. Everything else I was going to have to make up myself.

At the time I began working on the films, I actually had a little practice inventing Khuzdul words and names for the Middle-earth Roleplaying Game, to replace some earlier ones that weren’t quite Khuzdul-looking enough. Looking back at them, I can see there there were already some “philological puns” in there, some of which might be considered pretty atrocious. Some of them were based on well-known Khuzdul forms, such as those I’ve used before; but others borrowed inspiration from real-world languages, such as Old Norse and the Germanic languages ancestral to it.

For instance, one of the MERP words I was replacing was Khuzadrepa, intended (by its original creator, whose name I’ve never known, and who I hope, if he ever ends up reading this, is not offended) to mean “slayer of dwarves”. Obviously Khuzad was the creator’s attempt to create a combining form for Khazâd. But the second part, I believed, was not –repa but –drepa, the Old Norse word meaning “to kill.”

I liked the look of that, but I also knew that it could not be a Khuzdul form, given the absence of the sound p; but I turned it into a Khuzdul root DRF “kill”, and gave it the pattern of Tharkûn, producing darfûn, which I attached to khazad-, as a combining form already seen in Khazad-dûm: hence Khazad-darfûn. (Today I would probably have used udraf instead of darfûn.)

On a basis of which I can’t now be certain, I also created the root ʔZG “fight” (which I have not yet made use of again, but which I suppose still “exists” in some sense, since I haven’t decided that it’s not part of neo-Khuzdul) and created on the basis of the pattern seen in aglâb the word âzâg “fighting, battle” — where you can see that although the underlying form is *aʔzâg, the vowel and the glottal stop fuse to create a long vowel. This is a recurring feature of my version of Khuzdul — and since I don’t believe I’d yet studied Arabic at the time I was working on MERP, the similarity between this phonological rule and the Arabic one (which is basically the same) is perhaps “coincidental.” At any rate it seemed to me like an obvious development.

By the time I started work on The Lord of the Rings films, I had studied Arabic, so similarities of this type were thereafter usually pretty conscious.

The first Khuzdul translations I did for the films, way back in 1999, were very limited: they were a bunch of words that would appear in runes on the walls of Moria. I did not make much use of these translations later; indeed, I think I forgot about them for a long while — but they’re a good example of how I approached the translation. Here’s one of the examples:

Mabazgûn zai Azgâr Azanulbizarul zai shakâl Kheled-zâramul.
“Slain in the Battle of Azanulbizar on the shores of Kheled-zâram.”

I have to analyse this now almost as an outsider, lacking much (if any) direct memory of my thought-processes, and only advantaged by the fact that I have a general idea of how I think!

Mabazgûn is obviously “slain”, and shows a pattern maCaCC, similar to mazarb, and so I suppose a past participle. The ending –ûn is a little mysterious; perhaps it was intended to make the participle refer to a person, and was perhaps suggested by the –ûn in Tharkûn.

Zai must be a preposition, meaning “in/on.” I don’t think I’ve used it again.

Azgâr “battle” is pretty clear. It’s the same action-noun structure as in aglâb, and the root is ZGR, which is very obviously borrowed (in real terms) from the Adûnaic root of the same form. Could the Dwarves have borrowed such a root from speakers of Mannish languages? Or, conversely, could the Men have borrowed it from the Dwarves? I think the answer to both questions is yes — there’s nothing in Middle-earth history that I think would make it impossible.

It’s sometimes objected that Aulë made Khuzdul for the Dwarves, and he made it perfect, so it can’t have any borrowings. I don’t think this is actually stated by Tolkien — he does, indeed, identify kibil as a word with probable Elvish associations — and in any case I was not going to turn down a combination of sounds that seemed appropriate. In any case, several Adûnaic words made it into neo-Khuzdul, and the direction of the transmission is one of those things I’m content to leave lost in the mists of history — well, unless someone gives me a really good reason for interpreting things one way or the other.

Azanulbizarul and Kheled-zâramul are self-explanatory, I think.

Shakâl “shores” is evidently a plural, I suppose of shukl or shakl. I clearly wasn’t following any particular restrictions on the use of CaCâC plurals at this point.

Here’s another one:
Durin mabazgûn au Abzag Durinu
“Durin slain by Durin’s bane”

There are obvious objections to using the Mannish Durin in a Khuzdul context, especially when written in Dwarvish Angerthas. I was not and am not insensitive to the objections — which Tolkien had also noted, in the context of the appearance of the name of Durin on the West-gate of Moria. On the other hand, I felt (probably) that it was a bit above my pay grade to be inventing a “true Dwarvish name” for Durin the Deathless — which I still have not done. So Durin it remained.

Au must be a preposition meaning “by (an agent).” I haven’t used this one again either, as far as I know.

BZG is evidently a root for “slay, murder”, as in mabazgûn, while abzag is clearly a form derived from the same root, but with the pattern aCCaC. Clearly this is supposed to be related to uCCaC but different. I am sure that it’s not a casual error for “ubzag“; what I am not sure of is exactly what distinction I was making. Possibly it was that I imagined that “Durin’s Bane” was not conceived of as a person, and I associated the u– prefix with personhood.

It is, however, obvious why it is Abzag Durinu and not **Abzag Durinul: abzag literally means “killer” or “one who kills”, and therefore Durinu is in the “objective genitive” form.

There are a few more phrases here, but mostly short and not revealing a whole lot of grammatical information. In any case, the neo-Khuzdul seen here, though similar in general outline to that which I later developed, is obviously not exactly the same. When I come to frame something like a complete grammar of the language, I shall have to decide whether I want to retain any of these old forms and fit them in some how, or look at them as a mistake — or, perhaps, archaic “Moria-Dwarvish” forms which were somehow ancestral to the forms typically used in Erebor.

But there is an important point to be made here, about the way in which I had to work on neo-Khuzdul. I did not start with a complete grammar in mind, much less a complete vocabulary. I let the demands of particular translations build neo-Khuzdul up, bit by bit, until at last it developed some sort of coherence. Complete coherence I didn’t really expect and didn’t get, except that sort of coherence which can be imposed on a language thus developed after the fact.

It is odd to note that this is more or less the way in which Tolkien’s languages developed. Odd, not merely because I’m not Tolkien, but also because our patterns of work were quite different. Tolkien had years in which to carefully consider different aspects of his languages and carefully work out phonologies and inflectional paradigms. I had literally hours between the request for a translation and the time the end result needed to be delivered. Moreover, between one request and the next (particularly in less-used languages, like Khuzdul) months or even years might pass.

In the meantime, the likelihood that I would forget at least some parts earlier inventions was pretty good. Each time, I made an effort to, effectively re-learn the details of the languages I invented, in order to maintain consistency. Several things, however, might thwart this intention: first, I might not be able to find all of the older material in time to review it and get the translations done. Second, even if I found it, I might inadvertently overlook some details in the reviewing. Third, on reviewing the older material I might simply be unhappy with some aspects of the invention and choose to substitute something new.

Consistency was always my intention — but I wasn’t going to insist on an absolute consistency if the result was going to produce something that I thought wouldn’t really look like Khuzdul (or whatever other language I was working on). After all, natural languages are complex, inconsistent, and redundant, and a certain amount of synonymy or redundancy of construction could perhaps even enhance verisimilitude. I have often let myself be guided by my ear (and my gut), feeling that if I gave it free enough rein, the language would eventually speak for itself, and that if I made mistakes, they were not in failing to stick rigidly to an invention that I had made at one time, but in failing to listen closely enough to what the language was telling me about how it wanted to develop.

On the other hand, just playing by ear wasn’t going to be good enough if I didn’t want the language to descend into absolute chaos. So it was that when I was asked to translate lyrics into Khuzdul to be used in conjunction with the Moria scenes, I found the need — on translating complex verbal sentences for the first time — to sit down and create a verbal paradigm, which has remained essentially the same ever since.

Helge asks:

I take it that the words Ukhlat “grasper, holder” and Umraz “keeper” are based on the pattern seen in Tolkien’s uzbad “lord”, which is then taken as an agentive formation meaning *”ruler” or similar (*ZBD “rule”?) So assuming ZRB as the root meaning “write” (underlying the word Mazarbul), a “scribe” or “writer” would be *uzrab, if the theory holds?

That’s exactly right, and it brings me right to the next stage of the process of creating neo-Khuzdul: extracting every possible bit of meaning from the existing body of Khuzdul-vocabulary, and using it as a basis for further expansion.

As I’ve already shown, every Khuzdul word is the combination of a meaning-bearing root, and a syntactically or structurally significant pattern. Some of those patterns have already been seen:

This looks like a simple singular-plural pattern. But obviously things must be more complex: not all nouns have the CuCC pattern, and we have plurals of a different type, e.g.: bark “axe”, plural baruk “axes”.

This suggests that the CuCC/CaCâC type of pattern applies only to nouns of a specific class. Now, these classes could be arbitrary, like some declensional systems, or like the classes of Arabic “broken plurals”, in which case there would be nothing to do except to randomly assign new words to one class or another. But it also might be the case that there’s some connection between the form of the word and its semantics: in this case, considering the contents of the class, it might be that the CuCC/CaCâC pattern applies to animate or rational beings. Such a theory is reinforced by the plural form Sigin-tarâg “Longbeards” — the name of a tribe of Dwarves.

We can easily guess that sigin means “long” and tarâg means “beards” (though the reverse is not impossible). But we can’t assume that the normal form for “beards” is tarâg outside of this compound — since beards as such are neither animate nor rational, while a Longbeard (dwarf) is. Presumable a single Longbeard is a Sigin-turg, but a beard by itself might be a targ, or something else with the same TRG root but a different pattern.

Incidentally, this is another “philological jest” — Longbeards translates a Germanic word derived from *Langabardôs, Latinized as Langobardi — this was the name of a Germanic tribe who invaded Italy, and whose name was gradually corrupted into “Lombard.” An early mediæval text in Latin, relating the origin of this people (the Origo Gentis Langobardorum) says that on the occasion of a war between the Vandals and a tribe called the Winniles, the women of the Winniles came to battle with their long hair let down and arranged around their faces in the shape of beards. This being seen by the deity Godan (=Óðinn), he said (in apparent astonishment) Qui sunt isti longibarbæ? — “Who are those longbeards?” — and from this came their name. The Longbeard dwarves owe nothing to this tribe other than the name, but perhaps the myth influenced Tolkien’s idea that the Dwarf-women resembled (and were bearded like) the Dwarf-men.

Anyway, I decided that I would use CuCC/CaCâC for all words referring to peoples: hobbits, elves, trolls, and so forth. As it happened, the only word of this type that I needed to create was “elf” — which became fund, plural fanâd. This is a “jest” of my own, although one which makes good sense in terms of the history of Middle-earth. The Dwarves had arisen in the early years of the First Age, after the Elves but before Men. They were unknown to the Eldar until after they reached Beleriand and met them in the Eryd Luin; but the Dwarves must have met other Elves before they encountered either the Sindar or the Noldor, and these were most likely either Nandor or western groups of the Avari, who (at this early stage in their history) probably went by the name of *Pendi. At any rate, fund is clearly an adaptation of pend– to Khuzdul phonology, substituting f for the p that is absent in Khuzdul, and using the CuCC pattern for “incarnates.”

Other words provided different meanings and patterns. On Balin’s tomb in Moria, we find him described as Uzbad Khazaddûmu “Lord of Khazad-dûm.” Now, it’s possible that uzbad “lord” is just a word, incapable of further analysis. But obviously it would be very convenient for me if I could get more out of it. I assumed that “lord” actually meant “ruler”, and that therefore the sequence ZBD meant “rule, govern” and the pattern uCCaC was the normal form for an agent — that is, in relation to any verb, a noun of this form would mean “one who [verb]s”. So, as Helge says, if ZRB was the root for “write, record” then a writer — most probably a professional writer, a scribe — would be an *uzrab. This was a pattern that I made considerable use of.

This theory about the meaning of uzbad also helped me explain why it is Khazaddûmu and not *Khazaddûmul, using the adjectival or genitival suffix which occurs so often elsewhere. I assumed that -u was the ending used for an objective genitive, one that can be used when the noun modified has verbal force, and the modifying noun is, in a sense, its object: that is, if uzbad Khazaddûmu can be understood to mean “one who rules Khazad-dûm.”

Menelion asks:

I have a question about phonetics: Tolkien states that kh is an aspirated k sound. And that’s OK if we have a vowel after it (as in Khazad). But what will we do in the case of words like “rukhs”? This is an after-stress position without a vowel, so I find this quite difficult to pronounce correctly. Do you think the sound would somewhat become more plain and closer to k or, maybe, there will be a schwa between kh and s?

Although both of these possibilities are of course possible, and both deaspiration and epenthesis of a vowel in difficult clusters do occur in real-world languages, in this case I take the transcription at face value: that rukhs represents [ʀukʰs]. The distinction between [ks] and [kʰs] is that in the latter case there’s an audible expiration of breath, a voiceless period of transition between the two consonants. If you wanted, you could represent it as [ə˳] — a voiceless schwa — but it would be very short, it wouldn’t be a syllabic nucleus in itself, and so phonologically wouldn’t count as a vowel.

But this reminds me that I promised to say something about Dwarvish aspirates and the lack of [pʰ] — and the presence of [f]. Khuzdul is unusual among the languages of Third Age Middle-earth in having voiceless aspirated stops — though the loremasters of the Eldar postulated such stops (pʰ, tʰ, kʰ, kʷʰ) for earlier stages of the Elvish languages. They can also be found in Adûnaic, Tolkien’s other Semitic-influenced language. In all of these languages (Khuzdul, early Elvish, and Adûnaic) there is a three-way contrast in stops. A stop may be voiceless and unaspirated  (like p, t, k) — meaning that the period of voicelessness, when the vocal folds stop vibrating, can be very brief. If preceded and followed by a voiced vowel sound — in, say, the sequence apa — the vocal folds stop vibrating only for the brief moment that the lips come together to make the sound p.

Or a stop may be voiceless and aspirated (like pʰ, tʰ, kʰ). This means that the voicelessness continues into the following sound. In the sequence apʰa, the first part of the following vowel a is voiceless — effectively, whispered. The sound of h in English is a pure aspiration — it’s not so much a consonant in itself as it is the devoicing, or whispering, of the following vowel. If you look at yourself in the mirror while saying he, hay, ha, haw, hoe, who you’ll see that the h itself has no shape, and that the mouth immediately takes the shape of the vowel as you begin to say each word.

Or a stop may be voiced (like b, d, g). This means that the vocal folds continue to vibrate all the way through a sequence like aba.

Now the curious thing is that, although this kind of three-way voicing contrast was found in languages like ancient Greek and certain Indic dialects, or for that matter in ancient Chinese and Tibetan, it is absent from the Semitic languages. There indeed is a three-way contrast for some consonants in some Semitic languages; but the contrast is not based on voicing and aspiration, but on voicing and what is called, for a lack of a better term, “emphasis”.

What “emphatic” means with reference to Semitic languages depends in part on the language and in part on who is doing the analysis. They have been interpreted as velarized, pharyngealized, or glottalized or ejective sounds. What these all have in common is that they involve a certain amount of constriction of the throat coarticulated with the consonant. For our purposes, we don’t have to worry about the exact phonetic details, and we’ll just represent “emphasis” by that ambiguous symbol so much used by scholars of Semitic: the sub-dot.

The most common “emphatic” stops in Hebrew, Aramaic, and Arabic are ṭ (ṭêt in Hebrew, ṭā in Arabic) and ḳ, more commonly written q (ḳôp in Hebrew, ḳāf in Arabic). There is, notably, no emphatic p in these languages. The gap can be shown graphically this way:

This is rather distant from Khuzdul. There are a couple of things, however, which can bring it closer. One is that it appears that, in Hebrew and Aramaic at least, the voiceless stops may have been incidentally aspirated: that is, though phonologically /p t k/, phonetically they may have been [pʰ tʰ kʰ] rather than [p t k]. Two things point to this: one, that in Greek (which had the aspirated/unaspirated distinction) Semitic /p t k/ usually get transcribed as φ θ χ (pʰ tʰ kʰ) and not as π τ κ (p t k). In fact, τ and κ are usually used to represent ṭ and ḳ! The other is that in both Hebrew and Aramaic, /p t k/ in some environments (following a vowel, mostly) became fricatives [f θ x] (in some varieties of both Hebrew and Aramaic, even [f s x]) — which may suggest an intermediate stage of pʰ tʰ kʰ. In Yiddish, where the consonant values are derived from a rather late stage of Hebrew, the normal sound of כ (Hebrew /k/) is [x], while ט and ק (Hebrew ṭ and ḳ) are the normal representations of [t] and [k].

One might hazard a guess that Khuzdul — transcribed in alphabets invented by Elves — is no better represented than Hebrew transcribed into the Greek alphabet would be. If so, then possibly a table of Khuzdul stops would look something like this:

This makes Khuzdul look a lot more like Semitic, and explains why Khuzdul doesn’t have a p; it would be equivalent to an emphatic p, which is not found in the more common Semitic languages. But wait, there’s a problem – Khuzdul not only lacks p (emphatic p), it also lacks ph (/p/)! However, this is exactly the situation found in Arabic — where the phoneme cognate to /p/ in other Semitic languages is /f/. And we do indeed have an [f] in Khuzdul! If *p became f in Khuzdul just as in Arabic, then we could postulate a proto-Khuzdul consonant system substantially similar to that of Semitic.

This is all speculation, of course, and I have not the slightest shred of evidence that Tolkien ever actually thought along these lines. It is, however, a fairly neat way of justifying some of the curious asymmetries of the Khuzdul consonant system.