In the second chapter of the Silmarillion we learn that as soon as Aulë had made the Seven Fathers of the Dwarves, he "began to instruct the Dwarves in the tongue that he had devised for them". Their own name for their language was Khuzdul, which is evidently simply "Dwarvish", the Dwarves calling themselves Khazâd (singular Khuzd). We read that "according to their legends their begetter, Aulë the Vala, had made this [tongue] for them and had taught it to the Seven Fathers before they were laid to sleep until the time for their awakening should come. 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. But the change was so slow and the divergence so small that even in the Third Age converse between all Dwarves in their own tongue was easy. As they said, the change in Khuzdul as compared with the tongue of the Elves, and still more with those of Men, was 'like the weathering of hard rock compared with the melting of snow' " (PM:323).

Also Pengolodh comments upon "the tradition that they have...that Aule devised for them their tongue in its beginning, and therefore it changes little" (WJ:402). In contrast a gesture-language the Dwarves had devised for themselves, the so-called iglishmêk, was more changeable.

But however well-preserved, Khuzdul was rarely learnt by others than Dwarves themselves. Late legends had it that in Valinor, Aulë had acquainted Fëanor with the language he had made for the Dwarves, but Tolkien noted that this was not necessarily true; perhaps it was just a story due to the fame of Fëanor (VT39:10). In Middle-earth, the Elves were not particularly interested in Dwarvish, and they did not think highly of this language anyway: "They could understand no word of the tongue of the Naugrim [Dwarves], which to their ears was cumbrous and unlovely; and few ever of the Eldar have achieved the mastery of it" (Silmarillon ch. 10). Even Tolkien himself states as a fact that "Dwarvish was both complicated and cacophonous. Even early elvish philologists avoided it" (Letters:31).

But even when someone actually wanted to learn Khuzdul, the Dwarves themselves were very reluctant to teach it. Their own language was "a secret they did not not willingly unlock, even to their friends" (LotR Appendix F). One theory is that they felt that Khuzdul belonged exclusively to their own race, and that no others had any right to understand it. When they wanted to communicate with other races, usually for the purpose of trade, they would much rather learn the language of the others than teach them Khuzdul - even if the other party was willing to learn.

Only two or three times in all the long ages of Middle-earth did the Dwarves willingly teach their tongue to people of alien race. In the First Age, when the House of Hador first came into Beleriand from the east and met the Longbeards, a special friendship arose between the two races because these Men, being skilled riders, could offer the Dwarves some protection against Orcs. Then the Dwarves actually "were not unwilling to teach their own tongue to Men with whom they had special friendship, but Men found it difficult and were slow to learn more than isolated words, many of which they adapted and took into their own language." (PM:303; nonetheless, it seems that Khuzdul has influenced even the basic structure of Adûnaic, a language descended from the tongue of the early Edain.)

Elvish interest in Khuzdul was low in the First Age, but there was at least one exception: "Curufin was most interested in the alien language of the Dwarves, being the only one of the Ñoldor to win their friendship. It was from him that the loremasters obtained such knowledge as they could of the Khuzdûl." (PM:358) At least one Khuzdul word made its way into Sindarin: kheled "glass", that appears in Grey-elven as heledh (see the Silmarillion Appendix, entry khelek-). The Khuzdul word Khazâd, "Dwarves", was adapted to Quenya as Casar "Dwarf" and to Sindarin as Hadhod (the Dwarvish race being called Hadhodrim, WJ:388). Conversely, the Dwarves seem to have borrowed at least one word from Sindarin: kibil "silver" must be related to Grey-elven celeb.

Much later, in the Second Age, the Dwarves reluctantly allowed a few Elves to learn a little Khuzdul purely in the interest of science: "They understood and respected the disinterested desire for knowledge, and some of the later Ñoldorin loremasters were allowed to learn enough of both their lambe (aglâb) ["tongue" in Quenya and Khuzdul] and their iglishmêk [gesture-code] to understand their systems." Pengolodh the Loremaster of Gondolin is said "for a while to have dwelt among the Dwarves of Casarrondo (Khazad-dûm)" (WJ:395, 396). These later loremasters evidently had a less arrogant attitude than their colleagues in the previous age, who except for Curufin deliberately "avoided" Khuzdul (Letters:31).

On one point, however, the Dwarves were always "rigidly secretive... For reasons which neither Elves nor Men ever fully understood they would not reveal any personal names to people of other kin, nor later when they had acquired the arts of writing would allow them ever to be carved or written. They therefore took names by which they could be known to their allies in Mannish forms." (PM:304) Appendix F in LotR confirms this: "Their own secret and 'inner' names, their true names, the Dwarves have never revealed to any one of alien race. Not even on their tombs do they inscribe them." Hence the names Balin and Fundin, that occur in a Khuzdul context on the slab over Balin's tomb, are not themselves Khuzdul. They are Mannish names, merely the substitute names Balin and his father Fundin used when non-dwarves were present.

In chapter 20 of the Silmarillion, we are given one Dwarvish name, Azaghâl, the name of the Dwarf-lord of Belegost. Perhaps it is a title or nickname rather than his true "inner name". It has been suggested that it means "warrior", being related to the Númenorean verb azgarâ- "wage war" (SD:439). There is also the name Gamil Zirak, the name of a dwarf-smith, master of Telchar of Nogrod (UT:76). Perhaps it is just another nickname, or his name may have leaked to non-dwarves by accident, to his great and lasting regret. On the other hand, the Petty-Dwarves evidently did not attempt to hide their Khuzdul names. In chapter 21 of the Silmarillion, the Petty-Dwarf Mîm readily tells Túrin not only his own name, but also the names of his sons Khîm and Ibun. Perhaps such shocking indiscretion was one of the things the normal Dwarves hated the Petty-Dwarves for.

However, the Dwarves did not feel that it was improper to reveal the names of places. Gimli on his own initiative told the Fellowship what the Dwarves called the mountains over Moria and Moria itself: "I know them and their names, for under them lies Khazad-dûm, the Dwarrowdelf... Yonder stands Barazinbar, the Redhorn...and beyond him are Silvertine and Cloudyhead:...that we call Zirakzigil and Bundushathûr." (LotR1/II ch. 5) The Dwarves were not necessarily offended if others knew a few Khuzdul place-names.

When Gimli came to Lórien, still angry because the Elves at first required him to be blindfolded, Galadriel said to him: "Dark is the water of Kheled-zâram, and cold are the springs of Kibil-nâla, and fair were the many-pillared halls of Khazad-dûm in Elder Days before the fall of mighty kings beneath the stone." We are told that "the Dwarf, hearing the names given in his own ancient tongue, looked up and met her eyes; and it seemed to him that he looked suddenly into the heart of an enemy and saw there love and understanding. Wonder came into his face, and then he smiled in answer" (LotR1/II ch. 7). So Gimli perceived Galadriel's use of the ancient Khuzdul names as a friendly gesture. (Conversely: Back in the First Age, the Petty-Dwarf Mîm said of the hill he lived in that "Amon Rûdh is that hill called now, since the Elves changed all the names" - suggesting that this irritated him. The older, Dwarvish name of the hill was Sharbhund.)

Regarding Khuzdul, Tolkien stated that "this tongue has been sketched in some detail of structure, if with a very small vocabulary" (PM:300). It evidently came into being in the thirties. The Khuzdul names Khazaddûm and Gabilgathol turn up in an early Silmarillion version; see LR:274. Here we also find Khuzûd as the Dwarves' name for their own kind, later changed to Khazâd. The name Khazaddûm Tolkien first applied to Nogrod, not to Moria. Christopher Tolkien comments: "Khazaddûm is the first occurrence of the celebrated name. It is interesting to observe that it existed - but as the Dwarvish name of Nogrod - already at this time. Later the Dwarvish name of Nogrod was Tumunzahar... Gabilgathol, now first appearing, remained as the Dwarvish name of Belegost." (LR:278)

Of the Dwarvish language we are told that "structurally and grammatically it differed widely from all other languages of the West at that time" (PM:316-317). It seems that it was widely regarded as the proverbial "difficult language", like many Westerners think of Chinese today.

The phonology was in some respects peculiar compared to other contemporary languages. There were at least two aspirated stops, kh and th, that is, k and t followed by h. (Note that kh and th do not here denote German ach-Laut and th as in English thin, as these digraphs often do in Tolkien's spelling.) Initial English k and t are also aspirated, but probably not as strongly as in Khuzdul. Khuzdul also possesses unaspirated stops, like French and Russian k and t, but unlike the situation in both English, French and Russian, Khuzdul k and t are phonemes in their own right, that must be distinguished from kh and th. As we know very few Khuzdul words, it is hardly surprising that we have no minimal pairs, but k vs. kh and t vs. th are seen to contrast initially: Kibil-nâla vs. Khazad-dûm and Tumunzahar vs. Tharkûn. Other consonants include the voiced plosives b, d, and g, the unvoiced spirants f, s, and sh, the voiced spirants z and gh (unless the latter combination of letters is actually g + h), the lateral l, the vibrant r (some Dwarves used a uvular R, others evidently trilled R), the nasals n and m, the glottal h, and one semi-vowel, y. There was also a glottal stop, but this was possibly not a separate phoneme, just a weak sound that was prefixed to words with an initial vowel (see below).

The table of runes (Angerthas Moria) in LotR Appendix D includes two letters that are transcribed ps and ts; it is said that these runes were only used by the Dwarves. Is this a hint that these sounds were somehow significant in Khuzdul, maybe even functioning as phonemes in their own right?

According to VT48:24, Khuzdul does not tolerate "two initial consonants". All words must begin in a single consonant (or in a vowel). On the other hand, consonant clusters may occur at the end of words, as in Rukhs "Orc".

If some consonants were somewhat peculiar, the vowel system was pretty ordinary. The short vowels seem to form a classical five-vowel system, a, e, i, o, u, and they are all attested also as long vowels â, ê, î, ô and û. According to Tolkien's notes on the Angerthas runes, reduction vowels like the ones heard in butter were also common (and written as runes only used by Dwarves). However, such vowels are not directly attested (unless some of the u's and e's represent such vowels). Long vowels may be shortened when unstressed (?), compare Khazâd with Khazad-dûm. (Actually we know nothing about how Khuzdul words are accented; see, however, the entry salôn in the wordlist below for a possible hint.) It may also be that long vowels must be shortened before a consonant cluster; ûl "streams" becomes -ul- in the name Azanulbizar.

The basic structure of Khuzdul resembles that of Semitic languages, like Arabic and Hebrew. The stems from which words are derived are not by themselves pronounceable words, but consist of consonants only. Nouns, verbs, adjectives etc. are derived not only by prefixes and suffixes (if such devices are used at all), but also by inserting certain vowels between these consonants, sometimes also by doubling one of the consonants. Often the words are actually inflected by internal vowel-changes instead of adding affixes: Rukhs means "Orc", but plural "Orcs" is Rakhâs. The root consonants - the so-called radicals - remain the same, like *R-Kh-S in this case. In Khuzdul as well as in Semitic languages, there are usually three radicals in the root; several such roots are mentioned in TI:174 and RS:466: B-R-Z "red", B-N-D "head", K-B-L "silver", N-R-G "black". An example of a biconsonantal root is Z-N "dark, dim" (RS:466).

Obviously vowels will be added when these consonantal roots appear as actual words, e.g. baraz "red" or bund "head" from B-R-Z, B-N-D. The radicals *Kh-Z-D contain the general idea of "dwarvishness" and can be observed in such words as Khuzd, Khazâd "Dwarf, Dwarves" and Khuzdul "Dwarvish". The same radicals *Kh-Z-D are evidently present in the ancient Khuzdul name of Nargothrond, Nulukkhizdîn, but the precise meaning of this name is unknown (note that Nulukkizdîn in the Silmarillion ch. 21 is a misspelling; see WJ:180). The most basic meaning of Kh-Z-D may have something to do with the number "seven", compare Adûnaic hazid of this meaning (SD:428). The Dwarves were descended from Seven Fathers and were divided into Seven Kindreds - and as we know, dwarfs (sic!) are still associated with the number seven even in very late and very childish Mannish mythology.

The word uzn "dimness" provides an example of how a biconsonantal root (here Z-N) is treated. This word is apparently formed according to the same pattern as bund or Khuzd, but there is no initial consonant. LotR Appendix E refers "the clear or glottal beginning of a word with an initial vowel that appeared in Khuzdul": It may be, then, that a glottal stop does duty for the initial consonant in a word of such a shape. The glottal stop is apparently not represented in Tolkien's transcription of Khuzdul, but it had a special rune (# 35) in the Angerthas script.

As has already been mentioned, our Khuzdul corpus is very small. There are a few names, like Khazad-dûm and Zirak-zigil, the inscription on Balin's tomb, and a battle cry: Baruk Khazâd! Khazâd ai-mênu! "Axes of the Dwarves! The Dwarves are upon you!"

Baruk Khazâd! is said to mean "Axes of the Dwarves!" Baruk is usually taken to be an example of something similar to the Hebrew "construct state": the state a word is said to be in when it is placed in front of a noun to express a genitival relationship: X Y meaning "X of Y" or "Y's X". (Compare Hebrew sûs "horse", hammelekh "the king", sûs hammelekh "the king's horse, the horse of the king".) Of course, we cannot be certain that baruk is the normal plural "axes" and not a specialized form meaning "axes of". It may be significant that all the other attested plurals contain a long vowel: Khazâd "Dwarves", Rakhâs "Orcs", tarâg "beards", shathûr "clouds", ûl "streams", dûm "excavations, halls", bizâr "valleys". Could the normal plural "axes" be *barûk? Shathûr "clouds" may represent a plural pattern in -a-û-. In Hebrew, the vowels of words in the construct state are often shortened.

Or, given that u is clearly a Dwarvish element meaning "of" (Bund-u-shathûr "Head in/of Clouds", TI:174), is it incorporated in baruk, inserted between the second and third radical? Words with three simple radicals (1-2-3) seem to have singular forms in 1u23 (bund "head", Rukhs "Orc" - radicals B-N-D, *R-Kh-S) and plurals in 1a2â3 (Rakhâs "Orcs", compare Khazâd "Dwarves" and tarâg "beards" from *Kh-Z-D and *T-R-G). As baruk seems to have a similar radical structure of three consonants (*B-R-K), we may perhaps add a construct state plural in 1a2u3 to the paradigm and inflect B-R-K "axe" as follows: Singular *burk "axe", normal plural *barâk "axes", construct state plural baruk "[the] axes of" (and similarly e.g. *tarug Khazâd "the beards of the Dwarves" from the attested form tarâg "beards"?) The construct state singular may have the form 1u23u (*burku Khazâd "the axe of the Dwarves"), if Bundushathûr is simply *Bundu Shathûr "Head of Clouds" being written in one word when used as the name of a mountain (since B-N-D = "head").

The second part of the battle-cry is Khazâd ai-mênu! "The Dwarves are upon you!", our only real sentence. Ai-mênu is "upon you", ai being a short form of aya "upon" and mênu being accusative plural "you". This is evidently a nominal sentence, containing no actual Khuzdul equivalent of the verb "are". Sentences like this - "X Y" meaning "X is/are Y" - are common in Russian and many Semitic languages. This may support the theory of a distinct construct state of nouns, to distinguish "X Y" meaning "X of Y" from "X Y" meaning "X is Y".

Then there is the text that appears if one deciphers the runes on Balin's tomb: Balin Fundinul uzbad Khazaddûmu, "Balin son of Fundin, Lord of Moria." The names Balin and Fundin are Mannish, so their etymologies are irrelevant. What remains is the ending -ul, here used to form a patronymic, uzbad "lord" and the well-known name Khazad-dûm "Dwarrowdelf, Moria" (though there is no equivalent of the hyphen in the Runic inscription). It here occurs with an ending -u, that is evidently a genitive of some sort. But why is an ending required here when none is present in Baruk Khazâd "axes of the Dwarves"? (It does not matter whether baruk is a specialized form meaning "axes of" or is simply "axes"; even if it incorporates an element meaning "of", this inflection still affects the first word in the construction, not the second as in uzbad Khazaddûmu.) Evidently this is a kind of objective genitive, denoting that Moria is ruled by the lord, not simply that the lord somehow "owns" Moria (might that be *uzbud Khazaddûm, following the pattern of baruk Khazâd???) This theory finds strong support in Adûnaic, the Númenorean tongue, that descended from a Mannish tongue influenced by Khuzdul (SD:414). This language has a so-called "objective" form incorporating u that is used in compounds, e.g. gimlu-nitîr "kindler of a star" (gimlu- being the objective of gimli "star", SD:428 cf. 427). Though this Númenorean objective is used in compounds only and not independently as in uzbad Khazaddûmu, it may in origin be related to the Khuzdul objective.

The ending -ul is said to be a genitive ending of patronymics, but maybe it is not used to form patronymics only; see -ul in the wordlist below.

A few nouns are attested both in the singular and the plural: Khuzd pl. Khazâd, Rukhs "Orc", pl. Rakhâs (see also uzn in the wordlist below). As we speculated above, tarâg "beards" may be a plural formed according to the same pattern, so that the singular noun "beard" is *turg. The word shathûr "clouds" evidently belongs to another plural pattern than Khazâd and Rakhâs, and we cannot reconstruct the singular form. It would probably have the same radicals *Sh-Th-R, but different vowels. Other plural nouns are ûl "streams" and dûm "excavations, halls" (the latter may also be a collective). Is it significant that they both contain the same vowel û as shathûr? There is no such vowel in bizâr "valleys", though.

Only four verbs are attested: gunud "delve underground, excavate, tunnel" (stated to be a root), SLN "fall, descend swiftly" (VT48:24, only given as a triliteral base), felek "hew rock" and the related word felak, meaning to use a tool like a broad-bladed chisel, or small axe-head without haft. Felak may also be used as a noun denoting such a tool. Cf. English "hammer", noun or verb. This example indicates that Khuzdul verbs cannot always be distinguished from other parts of speech by their form alone.

We have a few adjectives: There is the word Khuzdul itself, apparently meaning "dwarvish", being derived from *Khuzd "dwarf" with the ending -ul that is also used to form patronymics: Fundinul, son of Fundin. We also have sigin "long" in Sigin-tarâg, the Longbeards. If Khuzdul adjectives agree in number, sigin may be a plural form. (On the other hand, the basic, uninflected form of the adjective may be used in compounds.) Zirak (pl. *zirik???) may be the adjective "silver" according to TI:174, but on the next page it is suggested that it means "spike" instead. It is possible that an adjective follows the noun it describes (though not in compounds like "Longbeards"); see below.

In compounds, the order of the elements is the same as in English: Khazad-dûm "Dwarrowdelf", Kibil-nâla "Silverlode", Kheled-zâram "Glass-lake" (concerning this translation rather than "Mirrormere", see wordlist), Gabilgathol "Great Fortress", Sigin-tarâg "Longbeards". The name Zirak-zigil "Silver-spike" (Celebdil, Silvertine) fits this pattern (TI:174), but Tolkien later seems to have decided that zirak means "spike" and zigil means "silver" rather than vice versa. In that case, this word may be a "construct state" connection just like baruk Khazâd seems to be: *Zirak zigil "Spike (of) silver" (a construction that Frodo, naturally ignorant of Khuzdul, took to be a compound and spelt Zirak-zigil, Zirakzigil). If zigil is an adjective "silver" rather than a noun, this construction may suggest that adjectives follow the noun they describe.

Only one pronoun is attested: mênu, plural accusative "you" (WR:20).

We have only two prepositions, aya "upon" (WR:20, reduced form ai in ai-mênu "upon you"), and u "in, of" (only attested in the middle of a compound, Bundushathur = "Head in/of Clouds", name of the mountain Cloudyhead, Sindarin Fanuidhol).

There is not much we can say about derivation. One derivational pattern seems to be of the form 1a23ûn, where 1, 2, 3 represent the three radicals. The meaning seems to be simply "person, thing or place characterized by the root meaning": Nargûn "Mordor, *Black Land", from the radicals N-R-G "black", and Tharkûn "Staff-man", Gandalf's Dwarvish name (radicals *Th-R-K "staff"?) If the consonants Z-Gh-L really are the radicals of the verb "to war" and Azaghâl means "warrior", we have an agentive pattern a1a2â3. The word Khuzdul "Dwarvish" may argue the existence of an adjectival pattern 1u23ul. But as stated above, -ul may be simply an adjectival ending added to the singular form of the noun (*khuzd "Dwarf"). Compare the patronymic Fundinul. If so, there is no need to establish a pattern 1u23ul that involves the original radicals.

Adjectives like baraz "red" (B-R-Z) or sigin "long" (*S-G-N) clearly represent adjectival patterns 1a2a3 and 1i2i3 (though kibil "silver" seems to be a noun).

The word Mazarbul, as in "the chamber of Mazarbul" (Chamber of Records), seems to represent some more complex derivation. If -ul is simply the adjectival ending discussed above (which would mean that the "of" in the translation is strictly superfluous), we are left with mazarb "record(s?)". Could this be a kind of passive participle, or the corresponding noun, of a verb "record" (radicals probably *Z-R-B)? If so, we have a pattern ma1a23.