On the one hand, Miyagawa (2012) reports that Japanese -mas- polite agreement marker may appear in embedded clauses headed by to (non-factive complementizer) and selected by say-type verbs (i.e. Hooper & Thompson's (1973) Type A verbs):
(i) Taroo-wa [Hanako-ga ki-mas-i-ta to] it-ta.
Taro-TOP Hanako-NOM come-POLITE-PAST C say-PAST
'Taro said that Hanako came.'
On the other hand, Japanese allows long-distance scrambling out of a clause headed by to:
(ii) [Sono hon-o]_i Hanako-ga [Taroo-ga t_i katta to] omotteiru.
that book-ACC Hanako-NOM Taro-NOM bought C think
'Hanako thinks Taro bought that book.'
My question is the following: In Japanese embedded -mas- clauses it's possible to legitimate an wh-phrase by a matrix interrogative ka particle like in (iii)?
(iii) Context: The speaker ask about Taro's words.
Taroo-wa [dare-ga ki-mas-i-ta to] it-ta ka?
Taro-TOP who-NOM come-POLITE-PAST C say-PAST Q
~ 'Who_i did Taro say t_i came?'
REFERENCES
Hooper, Joan B. & Thompson, Sandra A. (1973): On the applicability of root transformations. Linguistic Inquiry 4:465-497.
Miyagawa, Shigeru (2012) : Agreements that occur mainly in the main clause. In: Lobke Aelbrecht, Liliane Haegeman and Rachel Nye, Main Clause Phenomena: New Horizons, p. 79-111. Amsterdam; Philadelphia: John Benjamins.
Saito, M. (1992): Long distance scrambling in Japanese. Journal of East Asian Linguistics 1(1):69-118.