The Quenya dual form is a special plural used for when there are exactly two of the items in question: atanu “two men”, lasset “both leaves”. It can be translated as “two”, “both” or “(the) pair of”. It is used most frequently with items that are a “natural pair”, such as body parts:

• hendu “two eyes”, sg. hen (WJ/337).
• hlaru “both ears”, sg. hlas (PE17/77).
• mát “pair of hands”, sg. má (PE17/161; VT47/6).
• peu “the two lips”, sg. pé (VT39/11).
• piru “two toes”, sg. pirë (PE16/96).
• nápat “thumb and index finger (as a pair), [lit.] the pickers” (VT48/5).

Forming the dual: There are two separate dual inflections, -u and -t, with -u mostly used for consonantal nouns [hen (hend-) → hendu] and -t mostly used for vocalic nouns [lasse → lasset]. As indicated above, however, there are many cases where -u is used with vocalic nouns as well. As Tolkien describe it:

Original[ly] the Q. duals were (a) purely numerative (element ata) and pairs (element ū as seen in Aldūya); but they were normally in later Q. only usual with reference to natural pairs, and the choice of t or u [was] decided by euphony (e.g. ū was preferred after d/t in stem (Let/427, from a letter to Richard Jeffery written in 1972).

Thus the older proper dual suffix was -u. Tolkien indicated this original dual suffix was probably related to the root √YU “both” (PE21/74). It was anciently used with vocalic nouns by replacing the final vowel:

ū̆ also was little used in inflexional elements, save of the obsolescent dual, and that in AQ [Ancient Quenya] replaced and did not form diphthongs with the vocalic ending of noun-stems (PE19/106).

An ancient -u suffix was often preserved in frequently-used duals of vocalic nouns, like peu and piru above, as well as other examples like [ᴹQ.] ontaru “two parents”, sg. ontaro (Ety/ONO). Furthermore, Quenya disliked having two repeated dental stops, so -u was often used over -t in nouns containing t, d: aldu “two trees”, sg. alda (LotR/1110; MR/100).

The suffix -t was also the dual suffix for verbs (PE17/75; PE22/161), and it may have invaded noun formations in the same way that the plural suffix -r was copied from verbs to noun, though the invasion could have been in the reverse direction also. The t-dual is a more natural fit for most vocalic nouns, and is widely used for those, especially for less “natural” pairs: attat “two fathers” (VT48/19), ciryat “two ships” (Plotz), samat “two minds” (PE17/183). Tolkien elaborated on the primitive origin of both dual inflections at some length in a document on Common Eldarin: Noun Structure written in early 1950s:

Duality. This was shown by the addition of the elements u; s, th; ta (tta). The element ū was in most ways parallel to plural ī: it was always added direct to the stem and did not follow other inflexional elements; and it was originally employed only in nouns and pronouns, not verbs. But the elements s, th competed with it and could in CE be used in noun-declension as well as verbal, and could precede other affixes. ū, however, differed from ī in an important point: it did not form diphthongs with vocalic stems, but entirely replaced the vowel whether ómataima or vowel of suffix. This is a relic of the period when ū-duals were in fact separate derivatives of a base and not yet organized as parts of a declensional system. The elements ta, tta are clearly only reduction of the normal numeral element for “two”. Originally they could only appear in nouns, and were distinguished from the others in function, being a parallel to the partitive or indefinite plural: see above. That is: originally in CE appeared, say, talu “a pair of feet” of one person; but eleda’ta, eleda’tta “a couple of Elves” (PE21/73).

Although the two dual suffixes has distinct function in Common Eldarin, the two were blurred in Quenya so that “they were normally in later Q. only usual with reference to natural pairs, and the choice of t or u [was] decided by euphony” (Let/427, see above). Thus choice of t or u was driven mainly by the phonetic characteristic of the base word rather than any distinct meaning for the two suffixes (with the exception of some fossilized u-duals as noted above).

In various places Tolkien indicated that Quenya dual forms were “obsolescent” (PE19/106) or “archaic” (PE17/76). They were definitely part of Classical Quenya (Parmaquesta) from before the Exile, but may have fallen out of use in Exilic Quenya (Tarquesta). Most Neo-Quenya writers make use of the dual formation, however. One quirk of duality is that when nouns are explicitly numbered, they are mostly in the plural (eleni nelde = “three stars”) but two items remain in the singular: elen atta = “two star[s]” (VT49/45). The verb is also inflected in the dual to agree with the number, at least in older speech: nai elen atta siluvat “may two star[s] shine” (VT49/43-44), where siluvat = “shine-(future)-(dual)”.

Duals and noun cases: When combined with the various noun cases, the t-dual is very fusional, another sign of its archaic nature. The case inflection was attached directly to t-dual and then the combination was altered according to various ancient metatheses (consonant reversals). The following list of examples comes from the Plotz Letter as the various dual inflections of cirya “ship”; the ancient decompositions (marked with *) are my own extrapolations:

• ciryant [dual dative] = *cirya-t-n (with tn > nt).
• ciryato [dual genitive] = *cirya-t-o
• ciryanta [dual allative] = *cirya-t-na (with tn > nt).
• ciryalto [dual ablative] = *cirya-t-lo (with tl > lt).
• ciryatse [dual locative] = *cirya-t-se.
• ciryanten [dual instrumental] = *cirya-t-nen (with tn > nt).
• ciryatwa [dual possessive] = *cirya-t-wa.

The dual possessive form is not attested, but -twa is probably how it is inflected.

We don’t have any Late Quenya (1950+) examples of u-duals inflected for noun cases, but presumably they would be inflected as vocalic nouns, with the singular case inflection added to the dual form: hendunen = “with [by means of] both eyes” (hend-u-nen). There is one example from The Etymologies of the 1930s, however: ᴹQ. Veruen “of the Spouses” (Ety/LEP) the [ᴹQ.] genitive form of veru “husband and wife, married pair” (Ety/BES), since in the 1930s the genitive was formed using the suffix -n. Helge Fauskanger suggested in his Quenya Course (HFQC/Lesson 13) that perhaps the dative of a u-dual would likewise take the suffix -en: alduen “for two trees” (though Fauskanger did state this was a purely hypothetical construction). I think it is probably safer to assume the dative suffix -n is attached directly to u-dual form: aldun.

For nouns modified by a possessive pronoun suffix, the possessive suffix comes first, followed by the dual suffix: máryat “her (two) hands” (LotR/377; PE17/69) = má-rya-t, mántat “their hands (each both)” (PE17/161) = má-nta-t. Since the possessive pronoun suffixes always end in a vowel, presumably these possessed forms would use the t-dual inflections, even for consonantal nouns: atarinyat “both my fathers” = atar-inya-t; atarinyant “for both my fathers” = atar-inya-nt (dative). Likewise, the dual forms of independent pronouns are mostly the plural or singular forms with the dual inflection -t: met “we two (exclusive)”, wet “we two (inclusive)”, tyet “you two (familiar)”, let “you two (polite)”, the main exception being tu “them two” (VT49/51).

Conceptual Development: The concept of the Quenya dual dates all the way back to Early Qenya, but in the 1910s the dual suffix was -wi rather than -u (PE12/21), probably derived from the primitive root ᴱ√WI (QL/33). By the 1920s the dual suffix was often -qi (PE14/52, 76-77). The dual suffix -t (and variant -s) was used mainly in verb inflections (PE14/76), though Tolkien may have imagined it invading noun inflections in the latter half of the 1920s, for example: dual ᴱQ. mallet “two streets”, sg. malle (PE16/114).

Tolkien’s longest description of the dual in his early writings comes from notes attached to the typescript version of the Early Qenya Grammar of the 1920s, where the dual was presented as an archaic formation:

In addition to the construction of satto “both”, yúyo “two” (see numerals), the following remains of the dual are still used:

The words for “eye, ear, nostril, foot, leg, hand, arm” when used in the plural referring to the members of one person only are construed with a singular adj. and verb, and with the following special forms. Poetically the verb may be in dual.

eye	hen	pl. hendi	du. henqi
ear	unko	pl. (unqi) unkoli	du. unqi
nostril	nĕn	pl. nengi	du. nenqi*
foot	tál	pl. tăli	du. talqi
leg	pelko	pl. (pelqi) pelkoli	du. pelqi
hand	má	pl. máli	du. maqi
arm	ranko	pl. rankoli	du. ranqi

* The usual term for the “nose” of one person: noses of several people is expressed by súni, pl. of súne “nose”.

These are also declined differently to plurals.

• N[ominative] A[ccusative] -qi; G[enitive] -qint; D[ative] -qit.

Cf. -t, -s ending of dual verbs (PE14/76).

Most of the above examples reappear in the contemporaneous English-Qenya Dictionary (PE15/67-79). The Early Qenya Grammar also described the notion of Quenya dvandva compounds (using Sanskrit terminology). These are either combinations of two distinct nouns in the dual, or a dual form with a meaning semantically distinct from the singular:

Note also the (poetical†) combination groups (dvandva compounds) with this ending, also construed as singular ordinal.

• Like: “sun and moon” ránuringwi, †ahúrasilqi
• Also **“twins” yungwi
• “heaven and hell” †valmandui, -manqi
• **“husband & wife” veringwi
• **“parents” atarqi, older †attahwi, puyandui (puita- “beget”).
• (trousers, socks, boots, gloves: see dictionary)

** These are in common use (PE14/76-7).

Notes from the 1950s indicate the role of such dvandva compounds was much reduced in Tolkien’s later conception of Quenya, limited to a few specialized duals such as veru “husband and wife, married pair” (Ety/BES) and ontaru “parents”.

The elliptical dual where two different nouns are associated, and the pair is expressed by one only with dual inflexion also occurred: as ontărū “begetters”, parents, father and mother. (Note that ū replaced the vowel-suffix: ontărō “begetter”.) But this was uncommon. Usually a word applicable to both was used as [Common Eldarin] Galadū “the Two Trees”. The compounding dual was not used (PE21/73-74).

Tolkien explored Qenya noun declensions in a series of tables posthumously labeled Qenya Declensions with various versions composed from the 1920s through 1940s (PE16/111-115, PE21/42-54). There is a longer document labeled the Declension of Nouns written in the early 1930s marking the mid-point of this series of conceptual developments (PE21/1-41). Complete dual paradigms appear only in a subset of these documents (versions 3, 4, 6), but the dual is at least mentioned in all of them.

Version 1 (PE16/111): The first version still has only the 4-case system of the Early Qenya Grammar, and the dual forms were deleted, but they show the introduction of -s/-t duals into noun declensions. The dual suffixes used with the vocalic noun ondo are:

• Nominative, Accusative -s; Genitive -vint; Dative -vi.

Version 2 (PE16/112): The second version adds the instrumental case. It provides dual examples for both a vocalic noun telko and a consonantal noun pilin (pilind-). The suffixes used (excluding archaic forms) are:

Noun Class	N.	A.	D.	G.	Inst.
vocalic	-unt¹	-ut¹	-ur¹	-n	-inent
consonantal	-u	-u	-ur	—	—

¹ Final vowel replaced with -u in vocalic nouns. Since the only example noun ends in -o which has special treatment in later versions, it is hard to say if this was the general pattern.

Version 3 (PE16/113-115): The third version adds the allative, ablative, locative and comitative cases. Excluding archaic forms and minor variants:

Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.	Com.
vocalic	-nt²	-t²⁺³	-ur¹	-u¹	-inent	-ntas	-llut	-sset	-uhto¹
consonantal	-unt	-ut³	-ur	-u	-inent	-atas	-ullut	-isset	-uhto

¹ Final vowel replaced with -u in vocalic nouns.
² Final -o replaced with -u in vocalic nouns.
³ Nouns whose last consonant cluster contains t, d change final -t to -s.

Version 4 (PE21/1-41): The fourth version adds a possessive case, but it is identical to the genitive for the dual case. It removed the comitative case. This version also adds a massive number of noun classes, but these are mostly irrelevant for the dual. Trisyllabic vocalic nouns show prosodic lengthening for suffixes beginning with a single consonant (-met):

Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.
vocalic	-t²	-t²	-u¹	-hta	-met	-ntau	-llut	-sset
consonantal	-unt	-ut	-u	-uhta	-(u)met	-(un)tau	-(ul)lut	-(us)set

¹ Final vowel replaced with -u in vocalic nouns (in version 4, sometimes appended instead).
² Final -o replaced with -u in vocalic nouns.
The parenthetic consonantal suffixes often added only the short form with various assimilations.

Version 5a (PE21/42-45): Version 5a restored the comitative. The only dual declensions were for sanga, ilduma.

Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.	Com.
vocalic	-t	-t	-tar	-tu	-met	-ndet	-llut	-sset	-lte

Version 5b (PE21/46-49): Version 5b modified the instrumental and ablative forms, as well as providing a larger set of examples.

Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.	Com.
vocalic	-t	-t	-tar	-tu	-nwen	-ndet	-llot	-sset	-lte

Version 5c (PE21/50-52): Version 5c reverted the ablative form and covered even more noun classes, including some consonantal examples.

Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.	Com.
vocalic	-t	-t	-tar	-tu	-nwen	-ndet	-llut	-sset	-lte
consonantal	-(a)t	-(a)t	-(a)tar	-(a)tu	-anwen	-(an)det	-(al)lut	-(as)set	-alte

The parenthetic consonantal suffixes often added only the short form with various assimilations; the longer -andet, -allut, -asset forms are speculative on my part (not attested).

Version 6 (PE21/53-54): The sixth version labeled the accusative simply as the “base” form to which it was identical (in this and earlier versions). This version only had vocalic examples.

Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.	Com.
vocalic	-t	-u¹	-úno¹	-tos	-únen¹	-ntas	-llos	-sset	-úto

¹ Final vowel replaced with -u in vocalic nouns (in version 6, sometimes appended instead).

Putting all of the above into one big chart with “Version 0” for the Early Qenya Grammar and LQ for the Late Quenya paradigm from the Plotz letter:

V	Noun Class	N.	A.	D.	G.	Inst.	All.	Abl.	Loc.	Com.
0	vocalic	-qi	-qi	-qit	-qint	—	—	—	—	—
1	vocalic	-s	-s	-vi	-vint	—	—	—	—	—
2	vocalic	-unt¹	-ut¹	-ur¹	-n	-inent	—	—	—	—
2	consonantal	-u	-u	-ur	—	—	—	—	—	—
3	vocalic	-nt²	-t²⁺³	-ur¹	-u¹	-inent	-ntas	-llut	-sset	-uhto¹
3	consonantal	-unt	-ut³	-ur	-u	-inent	-atas	-ullut	-isset	-uhto
4	vocalic	-t²	-t²	-u¹	-hta	-met	-ntau	-llut	-sset	—
4	consonantal	-unt	-ut	-u	-uhta	-(u)met	-(un)tau	-(ul)lut	-(us)set	—
5a	vocalic	-t	-t	-tar	-tu	-met	-ndet	-llut	-sset	-lte
5b	vocalic	-t	-t	-tar	-tu	-nwen	-ndet	-llot	-sset	-lte
5c	vocalic	-t	-t	-tar	-tu	-nwen	-ndet	-llut	-sset	-lte
5c	consonantal	-(a)t	-(a)t	-(a)tar	-(a)tu	-anwen	-(an)det	-(al)lut	-(as)set	-alte
6	vocalic	-t	-u¹	-úno¹	-tos	-únen¹	-ntas	-llos	-sset	-úto
LQ	vocalic	-t	-t	-nt	-to	-nten	-nta	-lto	-tse	—

¹ Final vowel replaced with -u in vocalic nouns (in version 4 and 6, sometimes appended instead).
² Final -o replaced with -u in vocalic nouns.
³ Nouns whose last consonant cluster contains t, d change final -t to -s.
The parenthetic consonantal suffixes often added only the short form with various assimilations.

The occasional replacement of -t by -s in these declensions seems to be an earlier (1920s and 30s) solution to Quenya’s dislike of sequential dental stops, which in Late Quenya (1950+) was handled by choosing u-duals over t-duals.

We don’t have a lot of examples of how consonantal duals might have been formed in Tolkien’s later writings: the two best examples hendu “eyes” and hlaru “ears” are likely based on ancient “natural” u-dual forms, and don’t give much information on how duals would be formed for “ordinary” consonantal nouns. There is, however, a set of deleted “genitive honorifics”: -táro, -tárion, -táruo from base -tar in a set of notes from late 1950s or early 1960s (PE17/58). This implies (a) consonantal nouns use u-duals and (b) singular noun case forms are added to the u-dual. That’s the system I currently recommend until further evidence appears.

Neo-Quenya: Many other Neo-Quenya courses (including those of Helge Fauskanger, Thorsten Renk and Tamas Ferencz) suggest that consonantal nouns should also use t-duals with a joining vowel -e-: atanet “two men”, as opposed to my suggestion of atanu above. I believe this use of a joining vowel was originally popularized by Helge Fauskanger in his Quenya Course, and from there has infiltrated many Neo-Quenya documents.

There is no evidence that Tolkien ever used this formation, however, and I think it is more likely Quenya used the u-dual in such circumstances. In fairness to other Neo-Quenya authors, Tolkien did say the u-dual was originally used for “natural pairs” (Let/427, PE21/73) and most of the Late Quenya (1950+) u-dual examples fall into that category (body parts). I am of the opinion that this ancient distinction was mostly lost by the Parmaquesta/Tarquesta period. The consonantal inflections from the 1920s and 30s mostly use u-duals, with the exception of version 5c which seems to use a joining vowel -a-.