The basis for most relative pronouns is the root √YA. However, the most common relative pronoun is i, which is probably related to the definite article. The word i is frequently used to introduce a subordinate clause:

• Átaremma i ea han Ea “our Father who art in Heaven”.
• lá carita i hamil mára alasaila ná “not to do what you judge good (would be) unwise”.
• i carir quettar ómainen “those who form words with voices”.
• ar i Eru i or ilye mahalmar ea tennoio “and of the One who is above all thrones for ever”.

The relative pronoun i is very generic. It can refer to a person (“who”) or a thing (“what”), it can be the subject of the subordinate clause (“who art in Heaven”) or the object (“what you judge good”), and it can be singular (“who”) or plural (“those”). Though it superficially resembles the definite article, it is distinct from it, as shown in the fourth example which has both a definite article i “the” and a relative pronoun i “who”. It seems that this generic relative i is indeclinable, however, which is a strong indication that it is related to or derived from the definite article, which is likewise indeclinable.

In cases where the relative pronoun is declined, we generally see either ya (neuter) “what” or ye (personal) “who”, as in:

• yassen tintilar i eleni “wherein the stars tremble” (LotR/377), a locative plural referring to the “vaults of Varda”.
• yello camnelyes “from whom you received him” (VT47/21), ablative singular.

Tolkien discussed the personal (“who”) vs. impersonal (“what”) relative pronouns in late 1960s notes having to do with hands and fingers (VT47/21):

who rel. per. ye, pl. i
that rel. imp. ya

Probably these relative pronouns were developed from √YA in the same pattern as personal se vs. impersonal sa. Note that the plural form of ye was i, which could be declined as well: ablative plural illon or genitive ion (VT47/21). No plural form was given for ya, but it may be *yai (compare tai; hat-tip to Luinyelle for this suggestion). It is not entirely clear when you should use declinable ya/ye vs. indeclinable i. There is at least one example of ya used as a relative pronoun by itself, without any inflections:

• sí ar lúmesse ya firuvamme “*now and at the hour of our death” (VT43/28), more literally “now and at the hour that we will die”.

One interesting derivative of √YA is the (adverb?) yá “when”, used conjunctively in the sentence yá hríve mene, ringa ná “when winter comes, it is cold” (VT49/23). This could be a temporal relative adverb, analogous to demonstrative sí “now” and tá “then”. In the previous version of the sentence from VT43/28, Tolkien used yá instead of ya: sí ar lúmesse yá firuvamme (VT43/27) perhaps: “now and at the hour when we will die”. This is just a guess, though, since the variations of ya vs. yá could be driven by other factors like stress or Tolkien’s conceptual vacillations.

Origins of relative pronouns: Tolkien described the semantic development of the root √YA as follows:

ya-, used in Quenya as stem of relatives (being originally a demonstrative referring back to something behind, or previous in time) (PE17/66).

Indeed, there are also derivatives of this root having only to do with the past, such as yára “old”. But as a pronoun it seems to be used mainly as a relative pronoun, with the possible exception of ᴹQ. yana “that (the former)”. As noted above, personal ye “who” probably developed from impersonal ya by analogy with personal se vs. impersonal sa. As evidence of this, there is one place were ya might be used personally, hinting that ya was the default:

• yan i wilyar antar miquelis “to whom the air gives kisses” (PE16/96), with yan referring to a maiden fairy, though it is also possible that yan refers to the snowdrop which the fairy resembles.

As noted above, indeclinable relative pronoun i is probably derived from the same root as the definite article i. It is tempting to assume that it gained this function by analogy with the plural i of personal relative pronoun ye. However, in Sindarin i is both the article and relative pronoun as well, so I think it is more likely that this dual function was of more ancient origin.

Conceptual Development: In the Gnomish Lexicon (GL) of the 1910s Tolkien gave ī as the root of relative pronouns:

ī- root of relatives. as in í indefinite indeclinable relative particle (GL/50).

Such relative pronouns might have appeared in the contemporaneous text Sí Qente Feanor as suggested by Christopher Gilson (PE15/35):

• ᴱQ. varūse qentier nōvo san i malto īsier i nosta (PE15/32), possibly meaning “*in other manner than was said before by those from whom this birth was known”
• ᴱQ. nalto fustūme ma Melkon i var limpilto var tūkielto (PE15/32), possibly meaning “*they can be smelled out by Melko whom they are bound to or have looked for”

However, in the Early Qenya Grammar (EQG) of the 1920s Tolkien gave ya as the (indeclinable) relative pronoun:

The indeclinable relative pronoun is ya, which is either to be understood in any relation, or, very frequently, is defined by demonstrative or pronominal or adverbial words inside the relative clause (PE14/54).

This pronoun did appear later in the 1920s in a declined (r-dative) form:

• ᴱQ. yar i vilya anta miqilis “to whom the air gives kisses” (MC/215).

Both i and ya continued to appear thereafter, ya declined and i not, with the caveat that in the 1940s, i became in before vowels: “A clause in such cases is introduced by i, before vowels in (PE22/118)”. The personal form ye only appeared in Tolkien’s very late writings, and probably developed only after Tolkien abandoned masculine and feminine pronouns and switched to a distinction between animate (personal) and inanimate (neuter).

@@@ discuss sa vs. ha usage in QVS