Relative etc pronouns are made up of two words

As I mentioned a couple of times, I think the mi-conjugation (I should
say m-conjugation, since mean the secondary endings) was not in its
origin a conjugation, but that a set of non-finite forms, derived from
the bare verbal stem, which at that time served as a verbal noun or
participle, plus the deictic endings *-n,W "at me", *-s "at thee", *-t
"at him", and *-n "everywhere". Without these endings the verbal noun
alone is the ancestor of the "4th person", the t-less 3sg., with these
endings it is the ancestor of of 1sg, 2sg, 3sg and 3pl, respectively;
1pl, and 2pl of the m-conjugation are those of the hi-conjugation,
which was originally the only conjugation in PPIE.
In other words, if V is a verbal stem, then we have
*V "V-ing"
*V-n,W -> *V-m "V-ing at me"
*V-s "V-ing at thee"
*V-t "V-ing at him"
*V-n "V-ing everywhere"
In PPIE there were no conjunctions and no subordinate clauses. Instead
the function of subordinate clauses were filled with subordinate
constructions, as is the original state of affairs in the Uralic
languages. In those constructions, the above deictic participles or
verbal nouns could be construed with a subjective genitive in *-Vs and
an objective genitive in *-Vm, in the sequence SOV. This whole
participal phrase was, when suffixed (on the verb) with *-kWe/o,
formally an adverb. After people began to perceive this subordinate
construction as a subordinate clause and the deictic verbal noun /
participle as a finite verb (under the influence of Semitic
languages?) , they perceived the associated subjective genitive and
objective genitive as nominative and accusative, respectively, and in
order to distinguish them from the proper genitive function, they
separated those functions by root-stressing the two genitives when
functioning as nominative and accusative and end-stressing them when
functioning as 'proper' genitive, which became gen.sg. and gen.pl.,
respectively.


SOV languages, since they do everything the other way around, have
preposed subordinate clauses. PIE was a SOV language. Basque is a SOV
language. Therefore, for the better understanding of how SOV
languages like PIE handle their preposed subordinate clauses, it is
now time for cybalist's Basque lesson.

from R. L. Trask: The History of Basque
"
2.5 THE SUFFIX -ko
The 'relational' or 'adnominal' suffix -ko is of central importance in
the grammar of Basque, but it has often been misdescribed in grammar
books. Briefly, -ko can be added to virtually any kind of adverbial
phrase, regardless of its syntactic structure, to produce a complex
adjectival modifier which can appear within a noun phrase. The
adverbial phrase can be a lexical adverb, an adverb derived from an
N-bar, a case-inflected NP, a postpositional phrase, a participial
phrase involving an adverbial participle, an adverbial clause, or even
a complement clause. The only adverbials which cannot take -ko are
adverbs of manner derived from lexical adjectives (for which a -ko
derivative would carry no meaning beyond that of the lexical adjective
itself) and those for which a -ko derivative would be senseless. The
final -n of the locative and comitative cases is usually lost before
-ko, as is the -a of the locative singular. Examples:
atzo 'yesterday'
atzoko egunkaria 'yesterday's newspaper'
hemen 'here'
hemengo jendea 'the people here'
esku huts 'bare hand' (Ñ)
esku-huska 'bare-handed' (adv.)
esku-huskako pilota partida bat 'a game of bare-handed pilota'
mendian 'on the mountain'
mendiko etxeak 'the houses on the mountain'
mendietan 'in the mountains'
mendietako haitzuloak 'the caves in the mountains'
Bilbon 'in Bilbao'
Bilboko kaleak 'the streets of Bilbao'
mendira 'to the mountain'
mendirako bidea 'the road to the mountain'
gurekin 'with us'
gurekiko neskak 'the girls who are/were with us'
dirurik gabe 'without money'
dirurik gabeko ikasleak 'students without money'
zu bezala 'like you'
zu bezalako pertsona bat 'a person like you'
mathai gainean 'on top of the table'
mahai gaineko liburuak 'the books on top of the, table'
trumoiak adituta 'having heard the thunder'
trumoiak aditutako umeak 'the children who had heard the thunder'
guk ikusita '(having been) seen by us'
guk ikusitako jendea 'the people seen by us'
izarni agertu zitzaienean 'when the star appeared to them'
izarra agertu zitzaieneko garaian 'at the time when the star appeared
to them'
idi bat hegan ikusi zuela 'that he had seen an ox flying'
idi bat hegan ikusi zuelako kontua 'the story that he had seen an ox
flying'
...

The suffix -ko has a second function, formally quite distinct. In
certain circumstances it can be added to an N-bar to produce a
modifier. Examples:
hortz bi 'two teeth'
hortz biko sardea 'a two-pronged pitchfork'
bihotz on 'good heart'
bihotz oneko neska bat 'a good-hearted girl'
him urte 'three years'
hiru urteko ume bat 'a three-year-old child'
beso eder 'beautiful arm'
beso ederreko pilotaria 'a pilota player with a great arm'
The N-bar involved must be at least two words long, and there are some
ill-defined semantic restrictions.
There are a few other miscellaneous occurrences of -ko, a striking one
being the word balizko 'hypothetical', derived from the finite
verb-form balitz 'if it were', as in the old proverb Balizko olak
burdinarik ez 'A hypothetical forge doesn't produce any iron.'
"

As for the -ta of 'trumoiak adituta', note "Of great importance in
western varieties is the adverbial -ta participle, formed by adding
-ta to the perfective participle; ... ". The verb here, the so-called
radical or stem is adi- "hear", belongs to the class of verbs which
add the suffix -tu to form the perfect participle.


Basque does not use the -ko suffix for subordinate clauses, for which
is has other suffixes whicg are added directly onto the finite verb of
the subordinate sentence. In the following quote, note especially the
description of relative clauses.
"
2.10 SUBORDINATE CLAUSES
Unless it is very long, a subordinate clause quite commonly precedes a
main clause. The majority of subordinate clauses involve the addition
of a suffixed complementizer to the finite verb or auxiliary: either
-(e)la, which is [-wh], or -(e)n, which is [+wh]. These suffixes
produce a few phonological complications; among other things,
underlying forms of agreement suffixes show up overtly. The verb
bearing the complementizer typically comes last in its clause, but
need not do so, except in a relative clause, in which this is obligatory.
...
A relative clause takes -(e)n; the relative clause precedes its head,
the relativized NP is represented within the relative clause by a gap,
and there is no relative pronoun. The verb in the relative clause
agrees normally with the gapped NP:
Bihar hiria ikusiko dugu.
'Tomorrow we're going to see the city.'
[Bihar e ikusiko dugun] hiria oso zaharra da.
'The city we're going to see tomorrow is very old.'
Neskari musu bat eman diot.
'I gave the girl a kiss.'
Ez dakit [e musu bat eman diodan] neska nor den.
'I don't know who the girl I gave a kiss to is.'
Note that both diot and its relativized form diodan show agreement
with a third-singular indirect object.
Subjects (transitive and intransitive), direct objects and indirect
objects can all be relativized with complete freedom; these are the
three arguments with which the finite verb agrees.
"

Now compare this to the structure of the preposed Hittite relative
sentence.

from Sulvia Luraghi: Old Hittite Sentence Structure

305 (HG § 25 = A I 57.)
#a paprizzi kuis
he-makes impure who-NOM #
b III GIN KÙ.BABBAR pa:i
three shekel silver he-gives
#a Who makes it impure #b pays three shekels of silver.

2003 (HG § 28c = C III 4-5.)
#a n= an= zan pitte[nuz]zi kuis
conn she-acc refl he-kidnaps which-nom
#b n= an= si= k[an] tuhsanzi
conn she-acc him-dat ptc they-separate
#a The one who has eloped with her, #b they take her away from him.

2004 (HG § 43 = IV 15.)
#a suwayazi= ma= an kuis
he-hits conn him-acc which-nom
#b nu= za apun= pat da:i
conn refl that-acc ptc he-takes
#a The one who hits him, #b that very one he pursues [?].

One English translation of these sentences would start with 'He
who...'. Hittte being a SOV language has it backwards: 'who-he'. As I
proposed earlier the relative (like the demonstrative) pronoun is made
up of two words, PPIE *kWe/o, a relativizing particle which is added
to relative clauses, as in Basque (remember that the forms of the
mi-conjugation was not originally finite, they were instead non-finite
participles), and the deictic pronoun *i-/y-, known from Latin is,
(e)a, id. The univerbation of these two result means two vowels meet,
and which one prevails must be conditioned by the rules of sandhi
for vowel collisions.

So, originally it must have been
*papri-t-kW(e), is ... "the impure-making (one), he..."
*ma suwaya-t im kW(e), is ... "but the hitting-him (one), he..."



Torsten