Эта версия была раньше доступна на сайте: http://forodrim.org/daeron/md_khuzdul.html, но позже заменена окончательной версией и теперь этот текст в интернете доступен только на нашем сайте.

Khuzdul, the language of the Dwarves, is for natural reasons difficult to analyze, since it is very sparingly used, and virtually never outside of purely dwarvish business. Aulë the Vala created the language at the same time as he created the dwarves, and he made it hard and harsh just like its speakers. This stiffness in the language is probably a contributing cause for its looking practically the same now as when it was made (this, and its being employed mostly in rituals and archaic usage). For this reason there are also almost no dialectal variants of the language; Dwarves understand one another even if they come from quite different geographic areas, provided that they talk together in Khuzdul.

Khuzdul has the five classic vowels a, e, i, o, u. These exist in short and long versions, and the latter are written with a circumflex (^) above. The pronunciation of the vowels ought to have been similar to that in the languages of Elves and Men, but in a note on Daeron’s runes in Appendix E in The Lord of the Rings, Tolkien hints that there were also reduced mid central vowels (see below). If so, this is a point in common with, among other languages, Hebrew, with which Khuzdul exhibits many further similarities (see below under structure). Hebrew has rules for when such a vowel is pronounced and when it is silent, but it is always written (the sign is called schewá and is written as two vertically arranged dots below the preceding consonant).

Khuzdul words beginning in a vowel are probably pronounced with a fairly strong glottal stop — concerning this assumption, see below under structure.

The consonants are also in principle the same as in other languages, as well as their pronunciation, with a few exceptions: Some Dwarves pronounced r as an uvular, i.e. in the back of the mouth, like Orcs. This seems to be the only dialectal distinction in Khuzdul. Also the combination of a voiceless stop and a following <h> is not pronounced as a fricative, as in the languages of Elves, but as an aspirate. <th>, <kh> and probably also <ph> (there is no example) indeed designate phonemes different from those written <t>, <k> and <p>, but they are pronounced as stops accompanied by a strong breath (more or less as in backhand, outhouse). I therefore prefer the typography <th>, <kh> and <ph> to avoid confusion. The combination <gh>, on the other hand, was presumably a (voiced velar) fricative, like in Black Speech and Orkish, as nothing special is said of that combination in the context of Khuzdul. And that should go for the combination <sh> as well, which would then be pronounced as in English.

Concerning syllabic stress not much is said, but in Semitic languages the stress is often on the last syllable. This could well be valid for Khuzdul as well, except for a group of nouns (the 2nd declension below) that in their basic form have a long vowel in the first syllable and a short vowel in the second. Here it sounds most natural to stress the first syllable, and these nouns would then in pronunciation agree with the segholates, a very frequent type of nouns in Hebrew, with an <e> (Hebrew seghól) in both of its syllables, and the stress on the first one (e.g. mélek, "king”).

Khuzdul is, as said above, in many ways similar to primary world Semitic languages, and where I have lacked examples it has been easy to fill out my analysis on the analogy of Hebrew and its relatives. To begin with, the words (at least the nouns — other words are badly underrepresented in the available material) in Khuzdul are composed in the same way as the Hebrew verbs, sc. the stem consists solely of three (or two) consonants, that are supplemented with vowels in between to constitute words. These consonants are called radicals, and are common to all words with a similar meaning. An example is the radicals that designate things having to do with dwarves: Kh-Z-D. We see this stem in the attested words Khazâd, Khazad-dûm, Khuzd and Khuzdul (presumably also in the place-name Nulukkhizdîn). The word is fully defined only when vowels and any necessary endings or prefixes have been added between, after or before the stem radicals.

I mentioned before that my analysis to some extent is based on rules in the Semitic languages, and I will now deal with the most important similarities (attested and hypothetical) between them and Khuzdul, to lend some authority to my method in reconstructing Khuzdul noun paradigms, and also to explain how the paradigms are to be read.

Mid central vowel (”schewá”). In Hebrew only consonants were originally written, but later vowel signs began to be used. It became regular to attach vowel signs to all consonants except final ones (sometimes to them as well). The vowels were often fully pronounced, but in some positions they were reduced. The highest degree of reduction was ”schewá”, representing a lax mid central vowel, phonetic [ə]. Sometimes it was pronounced, sometimes not. Concerning Khuzdul, it is said in Appendix E of The Lord of the Rings: "vowels like those heard in English butter [i.e. mid back [ʌ] and mid central [ə]] . . . were frequent in Dwarvish and in the Westron”. They do actually have runes of their own in the certh table, reduced variants of which could be used when the vowels were "weak or evanescent” (Appendix E, II). But the runes for these sounds are not transcribed in the value table, and so were probably not pronounced enough to motivate a transcription.

The glottal stop (’alef). Both Hebrew and Arabic have a strong glottal stop in the beginning of words that in Latin transcription appear to begin in vowels. Its variants (voiced and unvoiced) have their own letters, and serve as radicals. Such a sound is said to have existed in Khuzdul as well. Some words that show only two consonants and begin in a vowel seem to be patterned in the same way as words with three radicals. Furthermore, this glottal stop might have been assimilated to a following consonant in the joint between two elements in a compound — see the special section about Nulukkhizdîn for more about this. Concerning the matter of a rune for this sound, it is said in the Appendices that one rune in the Angerthas Moria was used for "the clear or glottal beginning of a word with an initial vowel that appeared in Khuzdul” (Appendix E, II). The sound is there transcribed with ’ (an apostrophe). (Of additional interest in this connection is a theory that Scandinavians too used a specific rune for the glottal stop, or at least pronounced vowels in the beginning of words with a very clear glottal stop. This would possibly explain why Old Norse poetry alliterates words beginning in any vowel — what really alliterates is the glottal stop.) Throughout this article I have supplied the glottal stop, written as an apostrophe <’>, as a radical in words that seem to require it. Although Appendix E only speaks of the "glottal beginning of a word”, I expect that, just like in Semitic languages, this consonant may appear as first, second or third radical, though no example calls for a glottal stop as second radical.

Construct state. In Hebrew, possession is expressed in another manner than in the Indo-European languages. The word that is changed in form is the one signifying that which is possessed, not that which possesses, which in many European languages is inflected in the genitive case. In Hebrew the noun that is possessed is inflected in what is known as the construct state, and is followed by the possessing noun in its basic form, the absolute state. In Khuzdul there is at least one clear example of the same construction, namely the phrase baruk khazâd, "the axes of the dwarves”. Khazâd here stands in its normal attested plural form, the absolute state, while baruk is in the construct state. Here, though, Tolkien has stated that baruk is the same in construct as in absolute, so it is not visible on the word alone which is the case. Mostly, however, the construct state should shorten the vowels of the absolute, sometimes as far as to a schewá – it is common in Hebrew that vowels are reduced in the construct state, e.g. "bayith” ("house”) – "beth léhem”, ("house of bread; store”). In the charts I have included the four forms singular absolute, plural absolute, singular construct and plural construct.

Composition form. In Hebrew there is a so-called composition form, taken by nouns when they acquire a suffix (for instance, possessive pronouns are suffixes in Hebrew). It is often more or less identical to the construct state, and the difference is then indicated by context, or simply the presence / absence of a suffix. Khuzdul likewise seems to have had a specific form for nouns with an affix attached and in compounds, and it seems to have the same pattern as the construct forms. Therefore, I have here assumed that the forms are really the same, distinguished only by the context and the possible affix or compound, and I have used the term construct state for both. The composition form is evidently the form applied to the first element in attributive noun compounds, sometimes including a hyphen, for instance Khazad-dûm ("Dwarves-delving”). Incidentally the use of a hyphen is common in Semitic languages as well, but does not have quite the same function there as in Khuzdul.

I have identified a number of different types of nouns in Khuzdul, and from these examples I have hypothetically and analogically reconstructed five declensions. In the five declension charts below I show the pattern for each declension, using stems written with numerals in the place of radicals (e.g. 1 â 2 a 3). I give the forms singular and plural absolute (sing. / plur. abs.), and singular and plural construct (sing. / plur. constr.) with the objective form (obj.) appended where it is not simply a suffixed –u on the original pattern (see Affixes, -u). After each form I have given one or more attested or hypothetical examples of this form, the hypothetical ones being marked with an asterisk (*). Hypothetical here means "founded on logic”, which is normally quite spurious in natural languages, where psychology and chance rule. However, in Khuzdul logic ought to be applicable, as it is constructed by the master of structure and regularity: Aulë.

1st declension, type A (words with 3 radicals)

sing.abs.	1u23	e.g. rukhs (Orc), *khuzd (Dwarf)
plur.abs.	1a2â3	e.g. rakhâs (Orcs), khazâd (Dwarves)
sing.constr.	1u2a3	e.g. duban (valley of sthg)
plur.constr.	1a2a3	e.g. Khazad-dûm (Dwarves-delving / Dwarrowdelf)

1st declension, type B (words with 2 radicals)

sing.abs.	1û2	dûm/*tûm (delving / bold) [the d in dûm is assimilated by the preceding d in Khazad-dûm.],’ûl1 (streams) [glottal stop as 1st radical supplied by me];
obj.:	1u2u2	Tumunzahar (Hollowbold) [the last radical is assimilated m – n by the following dental].
plur.abs.	1a2â2	e.g. *tamâm (delvings / bolds)
sing.constr.	1u2	e.g. *’ul (streams of sthg)
plur.constr.	1a2a2	e.g. *tamam (delvings / bolds of sthg)

This type is originally the ordinary 1 u 2 3-pattern, but since the two last radicals are the same (*tumm) the vowel was prolonged and the radicals reduced to one (*tûm). This development is evident in Arabic.

2nd declension

sing.abs.	1â2a3	e.g. zâram (lake), nâla’ (way, (river)course(?)) [glottal stop as 3rd radical supplied by me], *zâbad (lord);
obj.	1u2u3	Buzundush (Blackroot), Nulukkhizdîn (Nargothrond) Further examples of this pattern may be the abandoned forms ’Udushinbar (later Bundushathûr) and ’Uruktharbun (possibly later Khazad-dûm).
plur.abs.	1a2û3	e.g. shathûr (clouds)
sing.constr.	1ə2a3	e.g. *zəbad (lord of sthg)
plur.constr.	1a2u3	e.g. *shathur (clouds of sthg)

3rd declension

sing.abs.	1a23	e.g. bark (axe),
plur.abs.	1a2u3	e.g. baruk (axes)
sing.constr.	1a23	e.g. ’aglâb (tongue, spoken language), Nargûn (black land / Mordor), Sharbhund (hill of the (bald) head (?) / Amon Rûdh) [subject to a metathesis, see The Metathesis], Tharkûn (staff-man / Gandalf)
plur.constr.	1a2u3	1 a 2 u 3  e.g. baruk (axes of sthg)

4th declension (i-declension — could possibly be applied with a diminutive function to words of other declensions) 

sing.abs.	1i2a3	e.g. zirak, (spike, peak)
plur.abs.	1i2â3	e.g. bizâr (streams)
sing.constr.	1i23	e.g. ’igl (language of sthg) [glottal stop as 1st radical supplied by me], Nulukkhizdîn (Nargothrond)
plur.constr.	1i2i3	e.g. *khizid ((Petty-)Dwarves of sthg) [Example taken from the early form of Nulukkhizdîn: Nulukhizidûn (see under Affixes, -ân, and the special section about this name).]

The plur. constr. is not attested, but here I have reconstructed a form inspired by the early name form Nulukhizidûn.

5th declension (words with 3 radicals; mass nouns without plural)

sing.abs.	1i2i3 alt.1e2e3	e.g. kibil (silver [the metal]), kheled (glass)
sing.constr.	1i2i3 alt.1e2e3	e.g. Kibil-nâla (Silverlode), Kheled-zâram (Mirrormere)

This is a special group of nouns that seem to denote various kinds of materials, and thus there is no (attested) plural form.

The two examples of this declension are reconstructed from the compound place-names Kibil-nâla and Kheled-zâram. I have assumed that the absolute form is the same as these attested construct ones.

There could very well be more declensions than these five. For example, there seems to exist a rather common pattern 1 â 2, in stated and hypothetical words like ’ân (river, from Gabilân, Sirion, [great river]; it is suggested in The War of the Jewels (p. 336) that there is a single element -ân meaning river), nâd (possibly (river-)course, used in an alternative name for Kibil-nâla’, and probably meaning the same thing as nâla’) and *’âb, from the word ’aglâb (tongue, spoken language). This could then be a sort of type B for the 3rd declension. However, nâd was never used in the final text, and the two others are treated as suffixes below, which I also deem them more likely to be.

Also we have the personal names Mîm and Khîm, which likewise could be the construct state of a hypothetical type B of the 4th declension. Sometimes other forms than the basic / nominative become names (e.g. the place-name Arboga in Sweden), so it is possible that these names could be related to hypothetical nouns like *mimam and *khimam, whose construct forms have given personal names to Petty-Dwarves, seemingly careless about revealing their inner names.

There are also patterns like 1 e 2 a 3, 1 a 2 o 3, etc. Concerning the latter, see Adjectives, gabil.

"1 NB: ’ûl (streams) is in The Lord of the Rings, The Reader’s Companion stated to be actually the adjectival ending -ul. The sole appearance of the word is in the placename Azanulbizar (note the composition construct state of the middle element in this triple compound, -’ul-, which would otherwise be -’ûl- — cf. the abandoned variant ’Azanûl), and concerning this name The Lord of the Rings, The Reader’s Companion disagrees with The Return of the Shadow. The latter states the name to be read "dim-streams-valley”, while the RC says that the valley-element (stated to be duban) is to be understood, and the name should be read "dimness-ish-streams”. I tend to support this hypothesis, since we then have a new stem explicitly described to mean "valley”, namely D-B-N, and the existence of this stem would be difficult to defend if there was another word having this meaning."

In Khuzdul it is, apart from the vowel variations, possible to alter a word’s meaning, and perhaps also its part of speech, by the use of suffixes and prefixes. Most of these affixes are hooked onto the construct form of the nouns, but there are also more autonomous affixes that can almost be seen as words of their own, or at least as fully productive suffixes/prefixes.

Below I have prepared an alphabetical list of the affixes of which I have found instances.

-âb   The meaning of this suffix is unclear; it could possibly signify a collective plural. That would then roughly go together with the plural pattern of the 1st declension. It is only attested in one place, in the word ’aglâb "spoken language” (glottal stop as 1st radical supplied by me). The radicals *’-G-L are recognized from the attested term ’iglishmêk (glottal stop as 1st radical supplied by me), the gesture-language of the Dwarves, and ought to signify "tongue/language”. The radicals Sh-M-K should accordingly have to do with signs or movements; regarding the form i 1 2 ê 3, see The Metathesis below. The difference in vowels between ’iglishmêk and ’aglâb makes us surmise that the 4th declension, the i-declension, can be applied to nouns of other declensions, and that the element ’igl derives from a word of the 3rd declension, *’agl.

(This element might alternatively be a noun of some declension beyond the five listed above, *’âb, see Nouns, the section after the last declension paradigm.)

-âl / -al   This is possibly an agentive ending added to verbs to denote someone who performs the action of the verb. This hypothesis is based on an interpretation of two names, Mahal and ’Azaghâl . The former is the Dwarves’ name for Aulë, not unlikely to mean simply ‘creator’. It is suggestive that the name is presented in the words "Aulë the Maker, whom they call Mahal” (The Silmarillion, ch. 2). This is one of only two places in The Silmarillion where Aulë is called "the Maker”.

If the latter, ’Azaghâl, is an agent noun, the stem *’-Z-Gh can possibly be the source of the Númenórean verb azgarâ ‘wage war’, and have a similar sense: either the ending -âl or the entire pattern 1 a 2 a 3-âl would then derive an agent noun ‘Warrior’ from this verb, a fitting (nick)name for the heroic Dwarf-lord it refers to. For verb stems of two radicals the pattern seems to be 1 a 2-al, without the circumflex, as in Mahal. That there is a short vowel in a shorter word indicates that there was a kind of vowel harmony rather than a vowel balance in Khuzdul. Vowel harmony is a phenomenon known from, among other languages, Finnish, a language whose influence on Tolkien’s creativity is familiar. Unfortunately this ending is not attested anywhere but in these two names, and verbs are on the whole badly under-represented in the existing material, so it is difficult to say more about the formation.

-ân / -în / -ûn This is a personifying ending, used to derive nouns from other nouns, and perhaps from practically any word. The attested instances are only names: Gabilân ("Great one”, Sirion – provided that -ân doesn’t mean "river” as suggested in The War of the Jewels (p. 336)), Nulukkhizdîn ("(Petty-)Dwarf-place upon the rivercourse”, Nargothrond), Tharkûn ("staff-man”, Gandalf) and Nargûn ("black land”, Mordor). ’Uruktharbun, an abandoned version of Khazad-dûm (probably), could also have this ending, but is not treated here since we have no indication to the meaning of the name. The ending itself must be translated to fit the context — "man”, "land”, or the like —, and the main words translate to different parts of speech ("(Petty-)Dwarf” and "staff” are nouns and "great” and "black” are adjectives). The variant -în may be the one regularly used on nouns of the 4th declension. An early form Nulukhizidûn was abandoned in favour of Nulukkhizdîn, showing yet another example of the presence of vowel harmony in Khuzdul, here applied on the quality of the vowel, rather than the quantity as was the case in the matter of Mahal (see -âl / -al above). Whether -ân or -ûn should be used in the other cases might depend on the type of the main word: -ân on stable adjectives like *gabil with two different vowels — which may, therefore, not be reduced —, and -ûn on nouns and reducible adjectives with two similar vowels (see Adjectives), which are reduced when taking this suffix. It may be added that for words with this suffix there is no plural attested, and it is quite likely that none exists, since place- and personal names seldom need number inflection. Neither should it be possible to apply this ending on a plural noun, since Nulukhizidûn was abandoned.

-în  See -ân.

ma-   Prefixed on a verbal stem, seemingly, but could actually be part of a full verb pattern ma-1 a 2 3. Probably forms perfect participle. The only attested example is Mazarbul, "belonging to the documented”. If this words ends with the suffix -ul "belonging to …” (see below) attached to a three-radical stem *Z-R-B, we are apparently left with a prefix ma-. As it happens, the Hebrew prefix for the perfect participles of derived verbs is mə- (e.g. məkhatev ‘written’), and it works well enough to assume a similar function for the Khuzdul ma-; *mazarb would be a (nominalized) perfect participle "[the] documented; records”, and *Z-R-B would denote keeping and documentation.

u-   Conjunction "and” (?).There is only one example of it (except the hypothetical example Udushinbar, see Adjectives, dush), but an example of unusual interest. It is the inscription on Balin’s sarcophagus: Balin Fundinul uzbad Khazad-dûmu – "Balin son of Fundin[, and] Lord of Moria”. The radicals in the word for "lord” ought to be *Z-B-D, and u- consequently a prefix. The full explanation might then be as follows: "lord” is in the construct state, although it is not really possessed by the following word, but rather positioned by it (due to the objective ending -u on Khazad-dûmu). If the word for "lord” belongs to the 2nd declension, and thus in its basic form is *zâbad, the phrase "Lord of Moria” would be *zəbad Khazad-dûmu, if it stood on its own. But now it is a further presentation of Balin; apart from being the son of Fundin, he was also Lord of Moria, and therefore a conjunction is fitting. Moreover, the word u- apparently reduces the vowel of the following syllable (if short to a schewá and if a schewá to nothing), while the second vowel remains, hence uzbad. These hypotheses I largely base on the Hebrew "wə-/u-" (without a glottal stop, but with a weak initial w-sound) This small word — it should be viewed as a word of its own — is most closely translated by "and”, but may often be regarded as a spoken comma. It is most likely more frequently used than the English conjunction — look up any chapter in the Old Testament and count the sentences beginning in "and”! The Hebrew ”u-” likewise has the power to reduce the vowel of the following syllable in some cases.

-u  This is an ending that gives the noun an objective or locative meaning. It is in some cases infixed rather than suffixed (see the noun paradigms above, the objective forms), but mostly it is simply suffixed without otherwise affecting the word. Although it sometimes makes the preceding word take the construct state, the meaning becomes objective rather than possessive, and is best translated as an objective genitive marker or a preposition.

In the phrase uzbad Khazad-dûmu from Balin’s tomb inscription, Balin is lord over or in Moria; sc. the ending -u on Khazad-dûm makes it an objective or locative (rather than possessive or adjectival) genitive attribute of the previous word, uzbad, which is apparently in the construct state (see above, u-). That the last word becomes -dûmu instead of *-dumum should be because the name Khazad-dûm is inflected as a whole and the connection to the original word *tûm is lost.

In Finrod’s Khuzdul epithet, Felakgundu "cave-hewer”, literally "hewer (of) cave”, gundu "cave” is likewise an objective genitive attribute of the noun felak, "hewer” (tool or person). The basic form thus ought to be *gund and go by the 1st declension. (This word may also appear in Gundabad, a name said to have a Khuzdul origin in The Peoples of Middle-Earth.) In this case it is harder to determine if the preceding word, felak, is in the construct or absolute state (though, according to the example above, it should be in the construct state), since it is the only attested noun of the form 1 e 2 a 3. It could possibly belong to a variant of the 4th ”i-declension”, but more likely it derives from the verb felek, "hew rock”, and thus conserves the [e] — even though the word felak itself has come to denote a verb: "to use a felak”. The name Bundushathûr could also contain this element, and in that case have the main words in reverse order: the suffixed word first, and the following noun is in the absolute state, not the construct, possibly because it is following instead of preceding. Divided into *bund-u shathûr it would then be "head-upon clouds”, or more understandably "clouds upon head”. (In The Treason of Isengard, p. 174, having said that bund(u) must be the part meaning ‘head’, Tolkien mentions "bund (B N D) - u - Shathûr ‘head in/of clouds’” as a potential analysis, though this could simply be an adaption to a more indo-european approach.) A last example of -u, again as an objective ending is perhaps found in the Dwarvish war-cry Khazâd ai-mênu, "The Dwarves [are] upon you”. mênu, ”you”, ends in -u, and is furthermore fairly obviously an object, of the prefixed preposition as well as the Dwarves.

-ul This ending forms adjectives, like the English ”-ish”. In the analysis of ’Azanulbizar, Tolkien identified it as "a genitive ending of patronymics such as Balin Fundinul” (The Lord of the Rings, The Reader’s Companion, p.269). This is a reference to the inscription on Balin’s tomb, where "Balin son of Fundin” is thus expressed by an adjectival genitive, more or less "the Fundinish Balin”. Since this second example shows that the ending can be attached to the Dwarves’ outer names, taken from the language of the Northmen, it should be possible to use it with any Khuzdul word. The word khuzdul itself apparently also contains this ending, and thus naturally enough would mean "Dwarvish”, "Dwarf-related”. In the name ’Azanulbizar, if literally meaning "dimness-ish-rills”, i.e. "[the valley of] dim rills”, as stated in The Lord of the Rings, The Reader’s Companion, the -ul- is in fact this adjectival ending suffixed on a plural of ’uzn (erroneously given as uzu): ’azan-, meaning "dimnesses”, and according to my theories a shortened (construct) form of the plural. This would mean two things: first that the adjectival ending -ul can be suffixed on a plural and furthermore the construct state (on the other hand we have the term khuzdul where it is suffixed to the absolute form), and second that the "valley” element is not written out – it is suggested in the same source that this element should then be duban, obviously a 1st declension noun *dubn in the construct state.

Also, we have this ending in mazarbul, "belonging to the documented” (see above, under ma-).

-ûn See -ân.

Generally, there seem to be two kinds of adjectives: Reducible and stable. Reducible ones resemble nouns in inflection, and mostly have two similar vowels which are subject to reduction, like the vowels in the noun patterns. Stable adjectives have a different pattern with mostly two different vowels, which are more stable in compounds and with affixes attached to them.

Comparative forms are not known at all. Number inflections of adjectives are seemingly not present, judging from the different translations of the two adjectives zigil and sigin as singular and plural, respectively.

The attested and hypothetical examples of adjectives are so few that it is possible to analyze each one of them separately.

[*’azan]  "Dim, dark”? Suggested in ’Azanulbizar, "dimrill (dale)”. This word is in braces, since I do not deem it as being an adjective. In The Lord of the Rings, A Reader’s Companion, it is said that azan is a plural noun meaning "dimnesses”, and was made an adjective only by the suffix -ul (see above, Affixes).

baraz "Red”. Attested in the name Barazinbar, "Redhorn”, Can obviously cause a metathesis of a following word, as other reducible adjectives, even when its vowels are not reduced, as in Barazinbar (see The Metathesis below).

*dush "Black”. Suggested in Buzundush, "Blackroot”, and in the abandoned Udushinbar for Bundushathûr "cloudy head”. In the name Buzundush, the element *dush would mean "black”, resembling the Orkish "dushgoi” describing Minas Morgul and implying that the word *dush would actually mean "dark” in Orkish as well. Udushinbar occurs only in the list Barazinbar, Zirakinbar, Udushinbar, and if this was meant to be the Dwarves’ traditional recitation of these mountain-names the last one might actually be *Dushinbar with the conjunction u- prefixed; in that case there is no basis for supplying an initial glottal stop.

*gabil "Great”. Suggested in Gabilân, "great one/river”, Sirion, and in Gabilgathol, "great fortress”, "Belegost”. There is probably a wordplay in the compound Gabilgathol; the word for "fortress”, gathol is almost identical to the Hebrew word for "great”: ”gadhol”. Here is also one of the two appearances of the vowel [o] in all the attested material, and the only one where the vowel is short; so if Tolkien had not made a little joke about the similarity between gathol and "gadhol”, short [o] might not have existed in the corpus at all.

gamil "Old”? From the name Gamil Zirak, "old spike(?)”. The meaning of this word is clearly indicated by the appearance of the parallel name "Zirak the old” in The Lost Road and Other Writings. Here we also have a possible wordplay in the similarity between gamil and poetic Old English ”gamol” (Modern Swedish ”gammal”), "old”.

narag "Black”. Attested in Narag-zâram, "black pool (?)” (abandoned name for Kheled-zâram), and Nargûn, "black land”, Mordor. Here is an example of two different forms with the same (adjectival) stem of radicals: 1 a 2 a 3 and 1 a 2 3. Obviously, when taking the -ûn-suffix, adjectives of the reducible type adopt the pattern of a 3rd declension noun. However this is not the case for the name Gabilân, which is stable, since it has two different vowels, and would then keep both vowels and take the -ân-ending.

sigin "Long”, pl. Attested in the Khuzdul name of Durin’s line: Longbeards (sigin tarâg). This is the only example of an adjective translated to a plural, but it is impossible to see from this example alone whether it is a special form for the plural or if Khuzdul has no number inflection of adjectives. However, the similarity to zigil, which is translated to a singular, indicates that there was no special plural form of adjectives.

*zahar "Hollow”. Suggested in the name Tumunzahar, "hollowbold”. The elements are here obviously in the reverse order as in English; the adjective thus comes last in this name like in Zirakzigil (see zigil). The noun, Tumun-, has an objective form (the original radicals of the noun should actually be *T-M-M, and the word should be from the 1st declension, type B; so the correct objective should then be *tumum, although the z in -zahar here assimilates the m to n), so it seems that when a noun precedes instead of follows its adjective, the noun takes the objective form. Why the noun in Zirakzigil does not is discussed below, under zigil, and under The Metathesis.

zigil "Silver (as colour)”. Attested in the place-name Zirakzigil, "silvertine”. This name was redefined by Tolkien from "silver-coloured + spike” to "spike + silver-coloured”. I hold the latter interpretation to be the more likely one, as it makes zirak into the noun, fitting the 4th declension pattern. zigil becomes the adjective, agreeing in pattern with the known adjective sigin, and also showing in the alternate name for Kibil-nâla’ : Zigilnâd, probably meaning roughly the same ("silverlode”). Also the translation of the name Gamil Zirak to "Old Spike” seems more logic to me than "Old Silver”, since "spike” could relate to a meaning of "on the top”, or "master”. The name Zirakzigil, like Tumunzahar, indicates the possibility to let the adjective follow the noun (without connecting them with a hyphen — the form Zirak-zigil was abandoned). Normally the noun in such a case should take the objective form (see above, *zahar), creating the name *Zirakuzigil. However, the much used shortened form of the mountain’s name, Zirak, may have affected the full name, and the objective form of the noun in that compound was abandoned.

The name Sharbhund appears to contain an adjective meaning "bald”, since the name is translated to Bald hill, but the truth should be that the name consists of two nouns, the second of which explicitly meaning "bald head”. The first noun would then mean "hill” and have the construct state, and the full name would mean "hill of the bald head” (see The Metathesis).

All this has referred to adjectives in attributive position, since predicative adjectives can only be attested in whole sentences, something that is lacking in the Khuzdul corpus. But since there was presumably no "be”-copula (as in Khazâd ai-mênu, "The Dwarves [are] upon you”), Khuzdul ought to have distinguished between attributive and predicative adjectives, in order to make a difference between phrases like ”the long beards” and ”the beards are long”. Possibly, predicative adjectives were formed with the apparently very productive adjective ending -ul (see Affixes above).

Verbs, often occupying the larger parts of grammars, do not even fill up a twentieth part of this account. The reason is of course the extremely sparse attested corpus, especially as regards full clauses. In all, there are three attested verbs, of which two are almost the same, or at least have the same base. The words are felek ("hew rock”), felak ("use a hewer, a felak”), and gunud ("make a tunnel”). There is also the lately attested stem S-L-N ("fall swiftly”); it is stated that "a name such as salôn or sulûn” would be a regular formation (Vinyar Tengwar 48:24). What the vowel patterns 1 a 2 ô 3 and 1 u 2 û 3 represent is not explained, except that we are dealing with noun-derivation (the possible origin of the name "Lhun”), and since they resemble no other pattern, it is very hard, if not impossible, to guess.

The analysis is difficult, not least because there are no examples of different forms of the same verb. Still, we may observe the word felak, which thus is on the one hand a noun (a kind of tool for hewers), on the other a verb (to use this tool). The root *F-L-K is obviously the same as in felek, so possibly a verb with two similar vowels could be made into a noun by changing one (or both?) of them. The noun would then make a separate "declension”, not fitting in any of the above given — the noun pattern 1 e 2 a 3 is unique for this word in the attested corpus. Further on, a noun can obviously be used as a verb, and so the root *F-L-K in its form felak may change back into a verb and mean "use a felak”. The fact that the nominal form can also be used as a verb indicates that verbs were identified by context and / or word order rather than by their isolated forms. The loose bounds between parts of speech make Khuzdul appear as a fairly simple and functional language in regard to verb clauses; this is also reasonable to imagine, in view of how and why the language was made.

In three attested compounds a metathesis seems to have arisen, changing the place of the first radical in the second element of the compound. These are Barazinbar, "Redhorn”, ’iglishmêk, the gesture-language of the Dwarves, and Sharbhund, "Bald hill”.

Barazinbar is composed of the adjective baraz, "red”, and the noun inbar, "horn”. The last word is seemingly not in its basic form. If one had only to simply put two words together to achieve a true compound word, this name should be *Baraznibar, provided that the noun with the radicals N-B-R (the attested radicals for the noun "horn” — Tolkien actually made a misprint when giving the radicals M-B-R) belongs to the 4th declension. But now there is a metathesis including the first radical and the first vowel of the second element, which produces the form Barazinbar.

The second example, ’iglishmêk, apparently has a metathesis including the same positions. It would with simpler rules be *’iglshimâk, if the second element is a noun of the 4th declension (and is here in its plural form), meaning "gestures”. A metathesis then results in *’iglishmâk. The vowel change [â] –> [ê] may be another evidence of vowel harmony, changing vowels in metatheses particularly.

Sharbhund differs from the other examples, since it seems to show a metathesis including only consonants, *sharhbund - sharbhund. Most likely the first noun is of the 3rd declension, absolute and construct state *sharh, and when connected to bund (here with a possible specific meaning: "bald head”), a metathesis takes place and *sharhbund becomes sharbhund, "hill of the bald head”.

About the cause of the metathesis, and the rule for when it appears, one can only guess. In the two latter examples, one could defend the metathesis by the difficulty to pronounce the consonant clusters *-glsh-, that appears due to the construct pattern 1 i 2 3, and *-rhb-, in the joint between *sharh- and -bund. Thus these examples suggest a rule that a situation of three radicals next to each other (arisen by reduced vowels in nouns and reducible adjectives) may provoke a metathesis of some kind, for the sake of pronounceability.

It is harder to formulate a rule that explains why Barazinbar has undergone a metathesis, and Gabilgathol has not, since the two words are rather similar in construction. But baraz, having two similar vowels, seems to be a reducible adjective, while Gabilgathol includes a stable adjective whose vowels are firm, and thus could cause no three-consonant cluster by reducing them. Even if baraz in this case does not reduce its vowels, it may have done so at first, creating the name *Barznibar which by metathesis became *Barzinbar. Then, inspired by the much used nickname for the mountain, Baraz, the Dwarves might have reconstructed the full name to Barazinbar, keeping the metathesis from the remembered form *Barzinbar.

The place-name Nulukkhizdîn (Nargothrond) gives a typical example of my work in reconstructing forms and grammatical rules in Khuzdul. I have in this case, as in many others, had good help from Jay Lawson, who first made me aware of the similar radicals of Nuluk- and nâla’. If these words really come from the same stem, one could come to the conclusion that a glottal stop might assimilate to a following consonant when appearing as the last radical of the first element in a compound. The glottal stop can indeed be assimilated to a following consonant in Arabic, so it would not be far-fetched.

The theory is further indicated by Tolkien’s early form of the place-name: Nulukhizidûn, where there is only one [k], which undoubtedly belongs to the second element. This early variant of the name, besides indicating that the suffixes -ân, -în and -ûn are really the same, also encourages my assumption that the plural absolute form of nouns from the 4th declension is 1 i 2 î 3, since this suffix is apparently by rule attached to a noun in the construct state, which mostly is the absolute form with shortened vowels. That this name was changed to Nulukkhizdîn may be explained by this assumption: 1 i 2 i 3 was a plural (construct state), and a suffix like -în would not be attachable to a plural, since it indicates a personal or geographical name.

We may thus recognize the radicals *N-L-’ from the noun nâla’ and the radicals *Kh-Z-D from the noun Khazâd. This leaves us with the question of the vowel-pattern 1 u 2 u 3 – 1 i 2 3 în. It has then given rise to my assumptions that the pattern 1 u 2 u 3 is the objective singular form of the 2nd declension noun nâla’, the pattern 1 i 2 3 is the singular construct state of the 4th declension noun *khizad, and the adapted form -în is the personifying ending -ûn here harmonized in vowel to the element -khizd.

Presuming that the first-mentioned stem *N-L-’ has the meaning "(river-)course”, and that the 4th declension indicates diminutive, the place-name Nulukkhizdîn would mean "Petty-Dwarf-land upon the rivercourse”. The nature around Nargothrond admits such an interpretation, since the mansions were located in the proximity of the river Narog.

What has been presented here is a body of largely quite hypothetical theories, or even logical conjectures, but even so I think I am reasonably qualified to conjecture, thanks to experience of created languages in addition to elementary university studies in Biblical Hebrew. In my search for attested instances I have been much helped by Helge Fauskanger’s webpage about Khuzdul on Ardalambion, and with respect to format and editing I owe great gratitude to Gildir, Per Lindberg, and Elros, Måns Björkman, Guild Master and secretary of Mellonath Daeron. The greater part of the translation into English, and much editorial recasting, was made by Beregond, Anders Stenström. Concerning many of the ideas and theories in this essay, especially the ones about Nulukkhizdîn and Tumunzahar, I have had much help from Jay Lawson, with whom I have had many and long discussions about this and Khuzdul in general.