A pseudo-sluicing analysis of reduced embedded questions in Chakhar Mongolian

Abstract

This paper provides a detailed description of reduced embedded questions in Chakhar Mongolian and proposes to analyze them in terms of so-called pseudo-sluicing. It has been noted in the literature that the comparable construction in Khalkha Mongolian does not exhibit the so-called case-matching effect, a phenomenon in which the case of a remnant interrogative phrase matches that of its correlate in the preceding context. We show that it also holds in Chakhar Mongolian. We argue that reduced embedded questions in the language have a pseudo-sluicing structure, which straightforwardly accounts for the absence of the relevant effect. Our proposal is supported by the appearance of a copula and pronominal subjects in those reduced questions and by the fact that reduced questions can be pragmatically controlled.

1 Introduction

The purpose of this paper is twofold. It aims to provide detailed descriptions for reduced embedded questions and related constructions in Chakhar Mongolian (henceforth, CM), which is the standard dialect of modern Mongolian spoken in the Inner Mongolia Autonomous Region, China. Reduction of interrogative clauses, or what is widely known as sluicing, has been studied in many languages (Merchant 2001; Merchant and Simpson 2012). As far as we know, however, the relevant phenomenon in CM has not been subject to close examination. This paper aims to add a new set of data from CM to the existing literature on sluicing. The other purpose of the present study is to propose an analysis of reduced embedded questions in CM that can account for the apparent absence of the so-called case-matching effect, first observed by Sakamoto (2012, 2015) for the comparable construction in Khalkha Mongolian (henceforth, KM), the standard dialect of modern Mongolian spoken in Mongolia. We argue that truncated interrogative clauses in CM are best analyzed in terms of what Merchant (2001) calls pseudo-sluicing, a structure consisting of a potentially null pronominal subject and a copula verb. We argue that remnant wh-phrases in reduced questions in CM are not case-marked precisely because they are complements of the copula verb.

The remainder of this paper is organized as follows. Section 2 presents brief illustrations of some basic syntactic characteristics of CM, setting the stage for the following discussions about reduced questions in the language. Section 3 succinctly reviews Sakamoto’s (2012, 2015) analysis of reduced questions in KM. Section 4 considers truncated interrogative clauses in CM in detail, ultimately pointing out that it is not remnant wh-phrases but the interrogative clauses containing them that are case-marked. Section 5 provides an analysis for reduced questions in CM, showing that it directly accounts for the absence of case-matching effect on remnant wh-phrases. Section 6 summarizes the entire discussion.

2 A profile of Chakhar Mongolian syntax

As in other dialects of Mongolian, the basic word order for simple transitive sentences is SOV in CM, as shown below.¹

(1)	Ene	xü-Ø	ene	nom-i	ungši-ba.
	this	boy-nom	this	book-acc	read-pst
	‘This boy read this book.’

The subject precedes the object, which in turn precedes the verb. The subject is assumed to bear nominative case, which is assumed to be zero morpheme in the language (Maki et al. 2015). The object is accompanied by the accusative marker i, which is alternatively realized as yi depending on whether it follows a consonant or a vowel.

Let us note that non-human indefinite objects are not marked with the overt accusative marker, according to Maki et al. (2015).² This is illustrated below.

(2)	Ene	xü-Ø	nom	ungši-ba.
	this	boy-nom	book	read-pst
	‘This boy read a book.’

When we have objects denoting human entities, they appear with the overt accusative marker irrespective of their definiteness (Maki et al. 2015).

(3)	a.	Mergen-Ø	nige	xümün-i	čoxi-ba.
		Mergen-nom	one	person-acc	hit-pst
		‘Mergen hit a person.’
	b.	Mergen-Ø	tere	xümün-i	čoxi-ba.
		Mergen-nom	that	person-acc	hit-pst
		‘Mergen hit that person.’

The object in (3a) is indefinite and the object in (3b) is definite. Both are accompanied by the overt accusative marker.

Like other dialects of Mongolian, CM is a wh-in-situ language (Janhunen 2012; Maki et al. 2015).

(4)	Ene	xeüxen-Ø	yaγu	uuγu-γsan	bui?
	this	girl-nom	what	drink-perf	prt
	‘What did this girl drink?’

In (4), the object is a wh-phrase, which stays in the object position in lieu of moving to the edge of the clause. Note that matrix questions in CM contain the particle bui, which appears in the final position.

Another major property of CM is that it is a pro-drop language, allowing arguments such as subjects and objects not to be overtly expressed. Consider the following data, where two speakers, A and B, engage in conversation:

(5)	A:	Batu-Ø	xen-i	olǰu	üǰe-gsen	bui?
		Batu-nom	who-acc	aux	see-perf	prt
		‘Who did Batu see?’
	B:	e	Suruna-yi	olǰu	üǰe-be.
			Suruna-acc	aux	see-pst
		‘lit. e saw Suruna.’
(6)	A:	Xen-Ø	Suruna-yi	olǰu	üǰe-gsen	bui?
		who-nom	Suruna-acc	aux	see-perf	prt
		‘Who saw Suruna?’
	B:	Batu-Ø	e	olǰu	üǰe-be.
		Batu-nom		aux	see-pst
		‘lit. Batu saw e.’
(7)	A:	Batu-Ø	Suruna-yi	olǰu	üǰe-gsen	üü?
		Batu-nom	Suruna-acc	aux	see-perf	prt
		‘Did Batu see Suruna?’
	B:	e	e	olǰu	üǰe-be.
				aux	see-pst
		‘lit. e saw e.’

The subject in B’s utterance in (5) is not overtly expressed (null arguments are indicated with e) though it is clear in the context that it refers to the subject in A’s utterance. Similarly, the object in (6B) and the subject and the object in (7B) are null but the sentences are perfectly acceptable.

Let us turn our attention to complement clauses in CM (Aravind 2021; von Heusinger et al. 2011; Maki et al. 2015; Sakamoto 2012).

(8)	Batu-Ø	[Tana-Ø	amitan-u	xüriyeleng	dotura-xi	tere	bars-ača
	Batu-nom	Tana-nom	animal-gen	garden	inside-indic	that	tiger-abl
	ayu-χu]-yi	(ni)	mede-ne.
	fear-inf-acc	ppc	know-npst
	‘Batu knows that Tana fears that tiger in the zoo.’

When complement clauses are selected by certain predicates like mede ‘know,’ they are case-marked in CM. The relevant clause in (8), indicated with brackets, is accompanied by the accusative marker. We may say that the complement clause serves as the object of the matrix verb and hence is marked accusative.³

Note that in (8), the case-marked complement clause is optionally followed by the particle ni, which is called a personal possessive clitic (PPC) in the literature (Hashimoto 2004; Sakamoto 2012). As PPCs are frequently used with various functions in CM, they deserve some discussion. Consider the following examples:

(9)	a.	Öxin	degüü-Ø	mini	Xöxeχota-dü	saγu-daγ.
		girl	young-nom	1sg.ppc	Hohhot-dat	live-hbt
		‘My younger sister lives in Hohhot.’
	b.	Öxin	degüü-Ø	čini	χamiγa	saγu-daγ	bui?
		girl	young-nom	2sg.ppc	where	live-hbt	prt
		‘Where does your younger sister live?’
	c.	Öxin	degüü-Ø	ni	Xöxeχota-dü	saγu-daγ.
		girl	young-nom	3sg.ppc	Hohhot-dat	live-hbt
		‘His younger sister lives in Hohhot.’

The PPCs in (9a-b) serve to indicate the first person and the second person possessor, respectively, of the subject noun phrases. The expression ni in (9c) is the third person possessive marker.

Some additional functions of the third person PPC are worth noting, as illustrated below (Sechenbaatar 2003; Sakamoto 2012; (10a-b) are cited from Gao 2014).

(10)	a.	Tere-Ø	ni	neite-yin	nom-un	sang
		that-nom	ppc	public-gen	book-gen	storeroom
		(bol-una).
		be-npst
		‘That is the public reading room.’
	b.	[Tere-Ø	öber-tegen	ire-xü]-Ø	ni	ǰoxistai.
		he-nom	self-dat.ref.poss	come-inf-nom	ppc	appropriate
		‘That he comes here himself is appropriate.’
(11)		[Mergen-ü	Begeǰing-dü	oči-γsan	čaγ]-Ø	ni
		Mergen-gen	Beijing-dat	go-perf.adn	time-nom	ppc
		öčögedür	bol-una.
		yesterday	be-npst
		‘When Mergen went to Beijing is yesterday.’

The particle ni can be used to indicate a third person nominal subject as in (10a) or a clausal subject as in (10b). In (11), which is a case of the pseudo-cleft construction, it is used to indicate the presuppositional clause (Sakamoto 2012).

What is noteworthy for the purpose of this paper is that ni appears after subordinate clauses. It is sometimes optional, especially with complement clauses, as in (8). In contrast, it is obligatory when following clausal subjects, as shown below.

(12)	a.	[Tere-Ø	kompani-du	ire-gsen]-Ø	*(ni)
		he-nom	company-dat	come-perf-nom	ppc
		nama-yi	soči-γa-ba.
		me-acc	surprise-caus-pst
		‘That he came to the company surprised me.’
	b.	[Man-u	anggi-yin	ali	nöxöd-Ø
		1pl-gen	class-gen	which	classmate-nom
		Begeǰing-dü	oči-χu]-Ø	*(ni)	todorχai
		Beijing-dat	go-inf-nom	ppc	clear
		ügei.
		not
		‘Which classmate in our class will go to Beijing is unclear.’

The bracketed parts here are clausal subjects. Our informants observed that the PPC ni is necessary in those cases.⁴

Embedded questions are also accompanied by ni.

(13)	a.	Tana-Ø	yaγuma	ǰigele-be.
		Tana-nom	thing	borrow-pst
		‘Tana borrowed a thing.’
	b.	Gebečü	bi-Ø	[Tana-Ø	yaγu	ǰigele-gsen]-i	(ni)	mede-xü
		but	I-nom	Tana-nom	what	borrow-perf-acc	ppc	know-inf
		ügei.
		not
		‘But I don’t know what Tana borrowed.’
	c.	Gebečü	bi-Ø	yaγu-yi	ni	mede-xü	ügei.
		but	I-nom	what-acc	ppc	know-inf	not
		‘But I don’t know what.’

Anteceded by (13a), (13b) contains an embedded interrogative clause, which is marked accusative and optionally followed by the PPC. In (13c), the embedded question is reduced to consist of the wh-phrase alone but is still accompanied by ni.⁵

3 Reduced embedded questions in Khalkha Mongolian

In this section, we look at reduced embedded questions in KM, reviewing Sakamoto’s (2012, 2015) observations and analyses. But before going into discussion about KM, let us note that Ross (1969) observes, based on data involving reduced questions in German and English, that the case of a wh-remnant must match the case of its correlate. Let us consider the German example below, cited from Ross 1969.

(14)	Er	will	jemandem	schmeicheln,	aber	sie	wissen	nicht
	he	wants	someone.dat	flatter	but	they	know	not
	*wen/wem.
	who.acc/who.dat
	‘He wants to flatter someone, but they don’t know who.’

The second clause here contains a reduced indirect question. The remnant wh-phrase must appear in dative case, just like its correlate jemandem ‘someone’ in the first clause. Though the main verb in the second clause, wissen ‘know,’ has the ability to assign accusative case to its nominal object, the remnant cannot appear in accusative case. This phenomenon is known as the case-matching effect (Merchant 2001; Abels and Dayal 2022).

Sakamoto (2012, 2015) observes that the case-matching effect is not present in reduced embedded questions in KM, based especially on (16) ((15) and (16) are cited from Sakamoto (2015)).

(15)	a.	Oyuna-Ø	yamar_negen_zuil-ig	zeel-sen.
		Oyuna-nom	something-acc	borrow-perf
		‘Oyuna borrowed something.’
	b.	Gevch,	bi-Ø	yu-g	n’
		but	I-nom	what-acc	ppc
		med-eh-gui.
		know-inf-neg
		‘But I don’t know what.’
(16)	a.	Bat-Ø	hen_negen-d	ene	nom-ig
		Bat-nom	someone-dat	this	book-acc
		ug-sun.
		give-perf
		‘Bat gave this book to someone.’
	b.	Gevch,	bi-Ø	*hen-d/hen-ig	n’
		but	I-nom	who-dat/who-acc	ppc
		med-eh-gui.
		know-inf-neg
		‘But I don’t know to whom.’

In (15), the correlate yamar_negen_zuil ‘something’ in (15a) is assigned accusative case, and the remnant yu ‘what’ in (15b) is also accompanied by the accusative case marker. This is consistent with the case-matching effect. Once we turn to data involving other cases, however, a different picture emerges. In (16), the correlate hen_negen ‘someone’ is assigned dative case in (16a), but the remnant wh-phrase appears with accusative case in (16b). If dative case were assigned to the remnant in order to be faithful to the case-matching effect, the result would be unacceptable, as shown above. Based on this and other observations, Sakamoto (2012, 2015) concludes that there is no case-matching effect for reduced embedded questions in KM.

To account for the obligatory presence of accusative case on remnants, Sakamoto (2015) proposes an LF-copying analysis of reduced embedded questions in KM (see Chung et al. 1995 for an elaborate exposition of the copying analysis). Specifically, Sakamoto (2015) proposes that remnant wh-phrases are base-generated in the specifier position of CP, the complement TP of which is empty, as illustrated below.

(17)	a.	[TP	Oyuna-Ø	yamar_negen_zuil-ig	zeel-sen].
			Oyuna-nom	something-acc	borrow-perf
		‘Oyuna borrowed something.’
	b.	Gevch,	bi-Ø	[CP	yu-g	[TP e]]	n’
		but	I-nom		what-acc		ppc
		med-eh-gui.
		know-inf-neg
		‘But I don’t know what.’
	c.	Gevch,	bi-Ø	[CP	yu-g	[TP	Oyuna-Ø
		but	I-nom		what-acc		Oyuna-nom
		e	zeel-sen]]	n’	med-eh-gui.
			borrow-perf	ppc	know-inf-neg
		‘But I don’t know what Oyuna borrowed.’

Anteceded by (17a), (17b) contains a reduced indirect question with the structure indicated. The wh-phrase is directly generated in the specifier position of CP and the embedded TP has empty structure. After the corresponding portion of (17a) is copied onto the empty TP, (17c) is obtained as the LF representation. As for the case of the wh-phrase in (17b), Sakamoto (2015) assumes that it is assigned accusative case by the matrix v, as indicated below (for the purpose of illustration, English glosses are used).

(18)

The idea seems to be plausible because the matrix v is associated with the transitive verb know and hence should have the ability to assign accusative case. It explains the presence of the accusative marker on the remnants in (15b) and especially (16b).

We have so far reviewed Sakamoto’s (2015) analysis of reduced wh-questions in KM. We will turn to CM in what follows. As will be shown, what Sakamoto (2012, 2015) observes for KM seems to be replicated in CM initially, but upon closer examination, a different analysis is necessary.

4 Reduced embedded questions in Chakhar Mongolian

Let us move on to consider CM, paying attention to how cases are assigned in reduced embedded questions. Like KM, CM seems to lack case-matching effects. Let us begin with the following data:

(19)	a.	Tana-Ø	yaγuma	ǰigele-be,
		Tana-nom	thing	borrow-pst
		‘Tana borrowed a thing,’
	b.	gebečü	bi-Ø	[Tana-Ø	yaγu	ǰigele-gsen]-i	(ni)	mede-xü
		but	I-nom	Tana-nom	what	borrow-perf-acc	ppc	know-inf
		ügei.
		not
		‘but I don’t know what Tana borrowed.’
	c.	gebečü	bi-Ø	yaγu-yi	ni	mede-xü	ügei.
		but	I-nom	what-acc	ppc	know-inf	not
		‘but I don’t know what.’

The sentence in (19a) is intended to antecede the full-fledged indirect question in (19b) and its reduced counterpart in (19c). The correlate yaγuma ‘thing’ in (19a) is an indefinite phrase and is bare (see (2) and note 1). In (19b), the wh-phrase yaγu ‘what’ is bare similarly. Note in passing that the indirect question is marked accusative with the overt accusative marker in (19b). Turning to (19c), we notice that the embedded question is reduced to consist only of a wh-phrase and that the wh-phrase is apparently accompanied by the overt accusative marker.⁶ Thus, we have a mismatch in case morphology between the correlate in (19a) and the remnant wh-phrase in (19c).

The observation made above is confirmed by another set of data.

(20)	a.	Batu-Ø	nige	xümün-dü	ene	nom-i	ög-be,
		Batu-nom	one	person-dat	this	book-acc	give-pst
		‘Batu gave this book to a person,’
	b.	gebečü	bi-Ø	[tere-Ø	xen-dü	ene	nom-i
		but	I-nom	he-nom	who-dat	this	book-acc
		öggü-gsen]-i	(ni)	mede-xü	ügei.
		give-perf-acc	ppc	know-inf	not
		‘but I don’t know to whom he gave this book.’
	c.	gebečü	bi-Ø	xen-i	ni	mede-xü	ügei.
		but	I-nom	who-acc	ppc	know-inf	not
		‘but I don’t know to whom.’

In (20a), which is intended to antecede (20b-c), the indirect object nige xümün ‘one person’ is marked dative. (20b) contains a full-fledged embedded question, where the wh-phrase xen ‘who,’ which corresponds to the indirect object in (20a), is marked dative as well. Once the embedded question is truncated as in (20c), however, the dative case disappears. Instead, the remnant wh-phrase appears to be accompanied by the accusative marker.

Let us add one more set of data, where the correlate is nominative but the remnant wh-phrase is marked accusative.

(21)	a.	Nige	xümün-Ø	Tana-yi	olǰu	üǰe-be,
		one	person-nom	Tana-acc	aux	see-pst
		‘A person saw Tana,’
	b.	getele	bi-Ø	[xen-Ø	Tana-yi	olǰu	üǰe-gsen]-yi	(ni)
		but	I-nom	who-nom	Tana-acc	aux	see-perf-acc	ppc
		mede-xü	ügei.
		know-inf	not
		‘but I don’t know who saw Tana.’
	c.	getele	bi-Ø	xen-i	ni	mede-xü	ügei.
		but	I-nom	who-acc	ppc	know-inf	not
		‘but I don’t know who.’

In (21a), the subject is an indefinite phrase marked nominative, which serves as the correlate of the wh-phrases in (21b-c). Note that while the wh-phrase in (21b) is nominative as expected, the remnant wh-phrase in (21c) is not marked nominative but appears with the accusative marker.

Note that according to Sakamoto (2015), remnant wh-phrases in KM are assigned accusative case by the matrix v selecting a transitive verb, as illustrated in (18). This leads to the expectation that if the predicates selecting reduced embedded questions are modified so that they can no longer assign accusative case, the remnant wh-phrases they contain should not be marked accusative.

One relevant case is obtained by passivizing the main verbs in (19c) and (20c). To examine it, let us first consider the active-passive alternation in CM.

(22)	a.	Bi-Ø	Tana-yin	sine	sondur-i	(ni)	toγu-ba.
		I-nom	Tana-gen	new	necklace-acc	ppc	like-pst
		‘I liked Tana’s new necklace.’
	b.	Tana-yin	sine	sondur-Ø	(ni)	nada-du	toγu-γda-ba.
		Tana-gen	new	necklace-nom	ppc	me-dat	like-pass-pst
		‘Tana’s new necklace was liked by me.’

In (22a), which is a typical active sentence, the agent argument is marked nominative and the theme argument appears with the accusative marker. When it is passivized, (22b) is obtained: the theme argument is promoted to the subject and marked nominative, optionally followed by the third person PPC, whereas the agent argument is demoted to the oblique (dative) phrase.

The same pattern is obtained in cases including clausal objects.

(23)	a.	Bi-Ø	[Batu-Ø	ene	χulaγaiči-yi	bari-γsan]-i
		I-nom	Batu-nom	this	thief-acc	catch-perf-acc
		(ni)	mede-ne.
		ppc	know-npst
		‘I know that Batu caught this thief.’
	b.	[Batu-Ø	ene	χulaγaiči-yi	bari-γsan]-Ø	ni
		Batu-nom	this	thief-acc	catch-perf-nom	ppc
		nada-du	mede-gde-be.
		me-dat	know-pass-pst
		‘That Batu caught this thief was known by me.’

In (23a), the matrix verb mede ‘know’ selects two arguments: the external argument is marked nominative and the internal argument, realized as a complement clause, is marked accusative. When the verb is passivized, (23b) is obtained. The complement clause does not bear accusative case but is marked nominative. Moreover, according to our informants, the PPC ni becomes obligatory in (23b) with the clausal subject, though it is optional in (23a) with the clausal object.

Bearing this in mind, let us proceed to reduced embedded questions to examine how remnant wh-phrases are case-marked when the embedded clauses are promoted to subjects by passivization.

(24)	a.	Mergen-Ø	yamar	nigen	xümün-i	toγu-na.
		Mergen-nom	some	one	person-acc	like-npst
		‘Mergen likes someone.’
	b.	Bi-Ø	[tere-Ø	xen-i	toγu-χu]-yi	(ni)
		I-nom	he-nom	who-acc	like-inf-acc	ppc
		mede-be.
		know-pst
		‘I knew who he likes.’
	c.	Bi-Ø	xen-i	ni	mede-be.
		I-nom	who-acc	ppc	know-pst
		‘I knew who.’
	d.	[Tere-Ø	xen-i	toγu-χu]-Ø	ni	nada-du
		he-nom	who-acc	like-inf-nom	ppc	me-dat
		mede-gde-be.
		know-pass-pst
		‘Who he likes was known by me.’
	e.	[Xen	*(bol-χu)]-Ø	ni	nada-du	mede-gde-be.
		who	be-inf-nom	ppc	me-dat	know-pass-pst
		‘Who was known by me.’

The sentence in (24a) is intended to antecede each of (24b-e). (24b) is an active sentence with an embedded question, which is marked accusative. Reduction of the embedded question in (24b) yields (24c), where the remnant wh-phrase is marked accusative. When we passivize the whole sentence in (24b), we obtain (24d), where the embedded question is promoted to the subject and marked nominative. If we truncate the clausal subject in (24d), (24e) is derived. In (24e), the main verb is passivized and hence, according to Sakamoto (2015), the remnant wh-phrase should no longer bear accusative case, which is borne out. Additionally, (24e) reveals something important: According to our informants, the reduced question cannot consist of the wh-phrase alone in (24e); the copula bol needs to be present. Further, they also reported that (24c) can optionally be realized as below.

(25)	Bi-Ø	[xen	bol-χu]-yi	ni	mede-be.
	I-nom	who	be-inf-acc	ppc	know-pst
	‘I knew who.’

Here the reduced embedded question consists of the wh-phrase with the accusative marker and the copula. The accusative marker follows the copula, which arguably means that the whole reduced clause is case marked. That (24c) can alternate with (25) suggests that they may be able to be analyzed in the following way:

(26)	Bi-Ø	[xen	(bol-χu)]-yi	ni	mede-be.
	I-nom	who	be-inf-acc	ppc	know-pst

The reduced embedded question here is comprised of the wh-phrase and the optional copula, and the accusative case is assigned to the clause, rather than to the wh-phrase. While it is necessary to account for why the copula is omissible only in clausal objects, the analysis in (26) is parallel to what is observed in (24e).

It is instrumental to look into the copula in CM at this point. The examples below are cited from Gao (2014).

(27)	a.	Ene-Ø	minu	degüü	(bol-una).
		this-nom	my	younger.brother	be-npst
		‘This is my younger brother.’
	b.	Tere-Ø	ni	neite-yin	nom-un	sang	(bol-una).
		that-nom	ppc	public-gen	book-gen	storeroom	be-npst
		‘That is the public reading room.’

The verb bol serves as a copula in (27) and can take nominal predicates. As can be seen in (27), the predicate nominals are not case-marked.⁷

Let us consider another set of data involving indirect questions in object and subject positions. The data in (28) consolidate what we have observed in (24).

(28)	a.	Mergen-Ø	nige	xümün-dü	nige-debter	nom
		Mergen-nom	one	person-dat	one-cl	book
		xürge-be.
		give-pst
		‘Mergen gave a book to a person.’
	b.	Bi-Ø	[tere-Ø	xen-dü	nige-debter	nom
		I-nom	he-nom	who-dat	one-cl	book
		xürge-gsen]-i	(ni)	mede-be.
		give-perf-acc	ppc	know-pst
		‘I knew to whom he gave a book.’
	c.	Bi-Ø	[xen	(bol-uγsan)]-i	ni	mede-be.
		I-nom	who	be-perf-acc	ppc	know-pst
		‘I knew who.’
	d.	[Tere-Ø	xen-dü	nige-debter	nom	xürge-gsen]-Ø
		he-nom	who-dat	one-cl	book	give-perf-nom
		ni	nada-du	mede-gde-be.
		ppc	me-dat	know-pass-pst
		‘To whom he gave a book was known by me.’
	e.	[Xen	*(bol-uγsan)]-Ø	ni	nada-du	mede-gde-be.
		who	be-perf-nom	ppc	me-dat	know-pass-pst
		‘Who was known by me.’

The sentence in (28a) serves as the antecedent for each of (28b-e). (28b) contains a full-fledged indirect question, which is assigned accusative case as the entire sentence is active. If the embedded question in (28b) undergoes reduction, (28c) is derived. Here, too, the copula is optional. When the copula is omitted, the remnant wh-phrase is immediately followed by the accusative marker, which is assigned to the entire embedded clause. If (28b) is passivized, (28d) is obtained. In (28d), the embedded clause is promoted to the subject and is marked nominative. If it undergoes reduction, it yields (28e), where the copula must be present and the wh-phrase is case-less. Comparing (28d) and (28e), we immediately notice that while the wh-phrase is marked dative in the former, it is not marked at all in the latter, arguably because it serves as the complement of the copula.

The reduced embedded questions in (24e) and (28e) are passivized subjects. Clausal subjects can appear in the causative construction as well.

(29)	a.	Batu-Ø	nige	xümün-dü	χairatai.
		Batu-nom	one	person-dat	fond
		‘Batu is fond of a person.’
	b.	[Tere-Ø	xen-dü	χairatai]-Ø	ni	nama-yi	sonirχa-γulu-na.
		he-nom	who-dat	fond-nom	ppc	me-acc	wonder-caus-npst
		‘Who he is fond of makes me wonder.’
	c.	[Xen	*(bol-χu)]-Ø	ni	nama-yi	sonirχa-γulu-na.
		who	be-inf-nom	ppc	me-acc	wonder-caus-npst
		‘Who makes me wonder.’

Anteceded by (29a), (29b-c) contain a full-fledged and a reduced clausal subject, respectively. Those subjects are marked nominative and followed by the PPC. The predicate χairatai ‘fond’ assigns dative case to its internal argument, as shown in (29a). The wh-phrase is marked dative as well in (29b). In contrast, the remnant wh-phrase is bare in (29c) and the copula must be overtly realized.

We have so far considered the cases where embedded questions are marked accusative or nominative. There are examples where they are case-marked differently.

(30)	A:	Man-u	kompani-yin	nige	xümün-Ø
		1pl-gen	company-gen	one	person-nom
		Xöxeχota-dü	aǰil-iyar	tomila-γda-na.
		Hohhot-dat	business-abl	assign-pass-npst
		‘A person in our company will be assigned to go on business in Hohhot.’
	B:	Bi-Ø	[Xen	??(bol-χu)]-du	??(ni)
		I-nom	who	be-inf-dat	ppc
		sanal	ügei.
		opinion	not
		‘I have no opinion on who.’
(31)	A:	Mergen-Ø	nige	xümün-dü	nom
		Mergen-nom	one	person-dat	book
		ög-čei.
		give-pst.con
		‘Mergen gave a book to a person.’
	B:	[Tere-Ø	xen-dü	nom	öggü-gsen]-eče
		he-nom	who-dat	book	give-perf-abl
		ni	aldara-γ.
		ppc	fade-mod.perm
		‘I don’t care to whom he gave a book.’
	B’:	[Xen	*(bol-χu)]-ača	ni	aldara-γ.
		who	be-inf-abl	ppc	fade-mod.perm
		‘I don’t care who.’

Here, speaker A’s utterances are intended to antecede speaker B’s utterances, which contain reduced embedded questions. The predicate used in (30B), sanal ügei ‘have no opinion,’ selects a dative complement clause, shown above. It is somewhat more difficult to omit the copula in (30B) compared with the cases where the reduced questions are marked accusative (see (28c)).⁸ The predicate aldara ‘fade’ in (31B) and (31B’) forms a fixed expression meaning ‘I don’t care’ together with an ablative phrase. While (31B) contains a full-fledged embedded question, (31B’) has a truncated clause, where the copula must be retained according to our informants. In (30B) and (31B’), the remnant wh-phrases are not case-marked but rather the entire embedded clauses are assigned dative case and ablative case, respectively.

Sakamoto (2012, 2015) does not consider those cases where reduced embedded questions serve as subjects or oblique phrases in KM. The data from CM given in (24e), (28e), (29c), (30B), and (31B’) do not seem to be amenable to Sakamoto’s (2015) analysis depicted in (18) because the reduced questions there contain the copula and bare remnant wh-phrases. According to Sakamoto (2015), truncated indirect questions consist of remnant wh-phrases in the specifier position of CP, which are case-marked by higher predicates, and empty TP, which is subject to LF copying. The data in question in CM indicate that case is assigned to the whole reduced interrogative clauses rather than to the remnant wh-phrases, because the case-markers actually follow the copula in (24e), (25), (28e), (30B), and (31B’), and that the copula appears obligatorily in some cases and optionally in others in allegedly empty TP. In the next section, we provide an alternative analysis for the phenomenon in question that can account for the presence of the copula and the (apparent) absence of the case-matching effect in CM.

5 A pseudo-sluicing analysis

Let us consider (24e) again. It is repeated as (32b) with its antecedent sentence in (24a), given here as (32a).

(32)	a.	Mergen-Ø	yamar	nigen	xümün-i	toγu-na.
		Mergen-nom	some	one	person-acc	like-npst
		‘Mergen likes someone.’
	b.	[Xen	bol-χu]-Ø	ni	nada-du	mede-gde-be.
		who	be-inf-nom	ppc	me-dat	know-pass-pst
		‘Who was known by me.’

The fact that the reduced embedded question in (32b) contains the copula leads us to assume that it has a so-called pseudo-sluicing structure (Merchant 2001; Adams and Tomioka 2012), exemplified below with English data.

(33)	a.	John bought something.
	b.	Guess [what it was].

The embedded question in (33b) contains a copula and a pronominal subject with the wh-phrase being the complement of the copula. Because CM is a pro-drop language as noted in Sect. 2, the embedded question in (32b) should be able to be analyzed as follows:

(34)	[pro	xen	bol-χu]-Ø
	she	who	be-inf-nom
	‘who she is’

Here the subject is a null pronoun referring to the correlate in (32a). The wh-phrase is the complement of the copula and hence is not assigned case [see (27)].

This automatically explains why CM apparently lacks the case-matching effect. Let us consider (20), repeated as (35).

(35)	a.	Batu-Ø	nige	xümün-dü	ene	nom-i	ög-be,
		Batu-nom	one	person-dat	this	book-acc	give-pst
		‘Batu gave this book to a person,’
	b.	gebečü	bi-Ø	[tere-Ø	xen-dü	ene	nom-i
		but	I-nom	he-nom	who-dat	this	book-acc
		öggü-gsen]-i	(ni)	mede-xü	ügei.
		give-perf-acc	ppc	know-inf	not
		‘but I don’t know to whom he gave this book.’
	c.	gebečü	bi-Ø	xen-i	ni	mede-xü	ügei.
		but	I-nom	who-acc	ppc	know-inf	not
		‘but I don’t know to whom.’

The correlate in (35a) is marked dative, and the corresponding wh-phrase in the full-fledged question in (35b) is marked dative as well. Once the question is reduced as in (35c), the wh-phrase seems to be assigned accusative case. We propose to analyze the reduced question in (35c) as follows:

(36)	[pro	xen	(bol-uγsan)]-i
	he	who	be-perf-acc
	‘who he was’

The clause contains a null pronominal subject referring to the correlate in (35a) (that is, the person to whom Batu gave this book) and the copula, which can be dropped because the clause is in the complement position of the verb mede ‘know.’ We assume that the accusative marker is attached to the clause, though if the copula is omitted, it turns out to be adjacent to the wh-phrase. The wh-phrase is the complement of the copula and hence is not case-marked. This explains why it does not show up with the expected dative case. More generally, according to the pseudo-sluicing analysis, remnant wh-phrases should be bare, which accounts for the lack of the case-matching effect in CM.

Note that the analysis shown in (34) and (36) predicts that null pronominal subjects in reduced questions in CM should be able to alternate with overt pronominal subjects. This prediction is borne out in the following data:

(37)	a.	Batu-Ø	nige	xümün-i	sigümǰile-be.
		Batu-nom	one	person-acc	reprimand-pst
		‘Batu reprimanded a person.’
	b.	Bi-Ø	[xen	(bol-uγsan)]-i	ni	mede-ye
		I-nom	who	be-perf-acc	ppc	know-1sg.imp
		geǰü	sana-ǰu	bai-na.
		that	hope-advl	aux-npst
		‘I hope to know who.’
	c.	Bi-Ø	[tere-Ø	ni	xen	(bol-uγsan)]-i
		I-nom	he-nom	ppc	who	be-perf-acc
		ni	mede-ye	geǰü	sana-ǰu	bai-na.
		ppc	know-1sg.imp	that	hope-advl	aux-npst
		‘I hope to know who he was.’
	d.	[Xen	*(bol-uγsan)]-Ø	ni	nada-du	mede-gde-be.
		who	be-perf-nom	ppc	me-dat	know-pass-pst
		‘Who was known by me.’
	e.	[Tere-Ø	ni	xen	*(bol-uγsan)]-Ø	ni
		he-nom	ppc	who	be-perf-nom	ppc
		nada-du	mede-gde-be.
		me-dat	know-pass-pst
		‘Who he was was known by me.’

The sentence in (37a) is intended to antecede each of (37b-e). In (37b), the reduced embedded question is in the complement position of the verb mede ‘know’ and hence the copula is optional. Significantly, in (37c), the reduced question contains an overt pronominal subject taking the correlate in (37a) as its antecedent. (37d-e) contain reduced questions as subjects. (37e) indicates that the overt pronominal subject can appear in the interrogative clause.

The pattern observed in (37) can be replicated with other wh-phrases.⁹

(38)	a.	Batu-Ø	marγasi	nige	γaǰar	xödelgegen-dü
		Batu-nom	tomorrow	one	place	event-dat
		orulča-na.
		attend-npst
		‘Batu will attend an event at a place tomorrow.’
	b.	Bi-Ø	[(tere-Ø	ni)	χamiγa	(bol-χu)]-yi
		I-nom	that-nom	ppc	where	be-inf-acc
		ni	mede-ne.
		ppc	know-npst
		‘I know where that is.’
	c.	[(Tere-Ø	ni)	χamiγa	*(bol-χu)]-Ø	ni
		that-nom	ppc	where	be-inf-nom	ppc
		nada-du	mede-gde-be.
		me-dat	know-pass-pst
		‘Where that is was known by me.’
(39)	a.	Batu-Ø	yamar	nigen	čaγ-tü	baγši-ača
		Batu-nom	some	one	time-dat	teacher-abl
		asaγulta	asaγu-ba.
		question	ask-pst
		‘Batu asked a teacher a question at some time.’
	b.	Bi-Ø	[(tere-Ø	ni)	xeǰiye	(bol-χu)]-yi
		I-nom	that-nom	ppc	when	be-inf-acc
		ni	γaiχa-ǰu	bai-na.
		ppc	wonder-advl	aux-npst
		‘I wonder when that is.’
	c.	[(Tere-Ø	ni)	xeǰiye	*(bol-χu)]-Ø	ni
		that-nom	ppc	when	be-inf-nom	ppc
		nama-yi	γaiχa-γulu-ǰu	bai-na.
		me-acc	wonder-caus-advl	aux-npst
		‘When that is makes me wonder.’

The sentences in (38a) and (39a) are intended to antecede (38b-c) and (39b-c), respectively. In (38b) and (39b), the reduced indirect questions are in the complement positions of the verb phrases. The wh-phrases may optionally be accompanied by the pronominal subjects and the copulas. In (38c) and (39c), the reduced questions are in subject positions. While the pronominal subjects are optional, the copulas cannot be omitted.

Additional evidence for the pseudo-sluicing analysis is obtained from the fact that reduced questions in CM can be used felicitously without linguistic antecedents. Before presenting relevant data in CM, we first note the dichotomy observed by Hankamer and Sag (1976) between ellipsis and pronouns (or more precisely, what they call surface and deep anaphora). They point out that while sluicing, which is assumed to involve ellipsis, requires verbally expressed antecedents, pronominal expressions can be used felicitously without such antecedents.

(40)	Hankamer:	Someone’s just been shot.
	Sag:	Yeah, I wonder who.
(41)	Context:	Hankamer produces a gun, points it offstage and fires, whereupon a scream is heard.
	Sag: #	Jesus, I wonder who.
(42)	Hankamer [observing Sag successfully ripping a phone book in half]:
		I don’t believe it.
(43)	Sag [same circumstance]:
		It’s not easy.

While (40) shows that sluicing is possible with a linguistic antecedent, (41) indicates that the mere presence of a context is not sufficient. In (42) and (43), on the other hand, the pronouns are used felicitously without verbally realized contexts.

Bearing these in mind, let us observe that reduced embedded questions in CM can be felicitously uttered without a linguistic antecedent (see Gribanova and Manetta 2016 for similar discussions on Uzbek). Consider the following examples (the context for (44) is modeled after Gribanova and Manetta 2016):

(44)	Context: Tana and the speaker are shopping in a boutique. The speaker picks up a mysterious product and says:
		Bi-Ø	[yaγu	(bol-χu)]-yi	ni	mede-xü	ügei.
		I-nom	what	be-inf-acc	ppc	know-inf	not
		‘I don’t know what.’
(45)	Context: The speaker hears someone screaming, and says:
		Eme	e,	bi-Ø	[xen	(bol-χu)]-yi	ni	mede-ye
		mother	prt	I-nom	who	be-inf-acc	ppc	know-1sg.imp
		geǰü	sana-ǰu	bai-na.
		that	hope-advl	aux-npst
		‘Oh my god, I hope to know who.’

Both (44) and (45) contain utterances with reduced indirect questions. Note that they are perfectly felicitous with the contexts given, which are not expressed linguistically. This fact supports our assumption that reduced questions in CM involve pronominal subjects.

It may be useful to mention at this point that reduced questions in CM do not exhibit island effects. As in other languages, relative clauses and adjunct clauses constitute islands for movement in CM (see Aravind 2021; Gong 2022 for related observations).

(46)	a.	Mergen-Ø	[Tana-du	nom	xürge-gsen]	xümün-i
		Mergen-nom	Tana-dat	book	give-perf.adn	person-acc
		olǰu	üǰe-be.
		aux	see-pst
		‘Mergen saw the person who gave Tana a book.’
	b.*	Tana-du	Mergen-Ø	[t	nom	xürge-gsen]
		Tana-dat	Mergen-nom		book	give-perf.adn
		xümün-i	olǰu	üǰe-be.
		person-acc	aux	see-pst
		‘lit. Tana, Mergen saw the person who gave a book.’
(47)	a.	Tana-Ø	[Batu-Ø	Mergen-i	čoxi-γsan	učir-ača]
		Tana-nom	Batu-nom	Mergen-acc	hit-perf	reason-abl
		uxila-ba.
		cry-pst
		‘Tana cried because Batu hit Mergen.’
	b.*	Mergen-i	Tana-Ø	[Batu-Ø	t	čoxi-γsan
		Mergen-acc	Tana-nom	Batu-nom		hit-perf
		učir-ača]	uxila-ba.
		reason-abl	cry-pst
		‘lit. Mergen, Tana cried because Batu hit.’

The example in (46a) contains a relative clause, shown with brackets. If the indirect object is extracted out of the relative clause by scrambling as in (46b), it results in an unacceptable sentence. The bracketed part in (47a) is an adverbial clause. The unacceptability of (47b), where the object is scrambled out of the adjunct, shows that it functions as an island in CM.

With these in mind, let us consider the following data:

(48)	a.	Mergen-Ø	[Tana-du	yaγuma	xürge-gsen]	xümün-i	olǰu
		Mergen-nom	Tana-dat	thing	give-perf.adn	person-acc	aux
		üǰe-be.
		see-pst
		‘Mergen saw the person who gave Tana a thing.’
	b.	Bi-Ø	yaγu-yi	ni	γaiχa-ǰu	bai-na.
		I-nom	what-acc	ppc	wonder-advl	aux-npst
		‘I wonder what.’
(49)	a.	Tana-Ø	[Batu-Ø	nige	xümün-i	čoxi-γsan	učir-ača]
		Tana-nom	Batu-nom	one	person-acc	hit-perf	reason-abl
		uxila-ba.
		cry-pst
		‘Tana cried because Batu hit a person,’
	b.	Bi-Ø	xen-i	ni	γaiχa-ǰu	bai-na.
		I-nom	who-acc	ppc	wonder-advl	aux-npst
		‘I wonder who.’

The sentences in (48b) and (49b) take (48a) and (49a), respectively, as their antecedents and contain reduced indirect questions. Note that the correlates of the wh-phrases occur inside the relative clause in (48a) and the adjunct clause in (49a). Our informants observed that the reduced questions are acceptable in the contexts given. This absence of island effects is compatible with our pseudo-sluicing analysis, which does not posit the structure containing islands for reduced questions.¹⁰

As a confirmation of involvement of pseudo-sluicing structure, we mention that (48b) and (49b) can optionally have pronominal subjects and copula verbs.

(50)	a.	Bi-Ø	[(tere-Ø	ni)	yaγu	(bol-χu)]-yi	ni	γaiχa-ǰu	bai-na.
		I-nom	that-nom	ppc	what	be-inf-acc	ppc	wonder-advl	aux-npst
		‘I wonder what that is.’
	b.	Bi-Ø	[(tere-Ø	ni)	xen	(bol-χu)]-yi	ni	γaiχa-ǰu	bai-na.
		I-nom	that-nom	ppc	who	be-inf-acc	ppc	wonder-advl	aux-npst
		‘I wonder who that is.’

The sentences in (50a-b) can be used after (48a) and (49a), respectively, without any problem.

Our pseudo-sluicing analysis would not be complete unless the optionality of the copula in reduced questions in the complement position is explained. Let us consider the following examples:

(51)	a.	Bi-Ø	[Mergen-Ø	suruγči	(bol-χu)]-yi	(ni)
		I-nom	Mergen-nom	student	be-inf-acc	ppc
		mede-ne.
		know-npst
		‘I know that Mergen is a student.’
	b.	[Mergen-Ø	suruγči	*(bol-χu)]-Ø	ni	nada-du
		Mergen-nom	student	be-inf-nom	ppc	1sg-dat
		mede-gde-ne.
		know-pass-npst
		‘That Mergen is a student is known to me.’
	c.	[Mergen-Ø	suruγči	??(bol-χu)]-yi	(ni)	bi-Ø
		Mergen-nom	student	be-inf-acc	ppc	I-nom
		mede-ne.
		know-npst
		‘lit. That Mergen is a student, I know.’

In (51a), the embedded clause is a propositional clause with the copula. Together with the optional PPC, it is adjacent to the verb, indicating that it is in the complement position of the matrix verb. (51a) allows the copula to be dropped optionally. (51b) is derived from (51a) through passivization. The embedded clause is promoted to the subject, and as a result, the copula cannot be omitted (as noted earlier, the PPC cannot be omitted, either). In (51c), the embedded clause in (51a) is dislocated presumably via scrambling, and it is somewhat difficult to omit the copula.¹¹, ¹² These examples show that, independently of reduced questions, the copula can be omitted only in embedded clauses in the complement position of verbs in CM.

The contrast between (51a) and (51b) is reminiscent of the possibility of that-omission in English.

(52)	a.	Everyone knows [CP (that) John is smart].
	b.	[CP *(That) John is smart] is known to everyone.

Stowell (1981) accounts for the impossibility of the empty complementizer in (52b) with the lexical government condition of the Empty Category Principle (Chomsky 1981): the empty complementizer must be governed by a lexical category. In (52a), it is the head of the CP complement of the verb knows, whereas in (52b), it is the head of the CP in the specifier position of TP (see Chomsky 1981; Stowell 1981 for more detailed discussions). (51a-b) may be explained similarly if we assume that they also involve omission of complementizers.¹³

Sakamoto and Bao (2019) argue that verbs raise to C via T in Mongolian. It is in part based on the possibility of so-called verb-echo answers like the one in (53B-B’).¹⁴

(53)	A:	Batu-bol	Baγatur-i	sigümjile-gsen	uu?
		Batu-top	Bagatur-acc	criticize-pst.adn	q
		‘Did Batu criticize Bagatur?’
	B:	Sigümjile-jai.
		criticize-pst.con
		‘lit. Criticized.’
	B‘:	[CP	Sigümjile-gsen	gejü]	bodu-na.
			criticize-pst.adn	c	think-npst.con
		‘lit. I think that criticized.’

Following Holmberg (2016) and Sato and Hayashi (2018), Sakamoto and Bao (2019) argue that verb-echo answers like (53B’) have V-to-T-to-C movement and TP-ellipsis, as shown below.

(54)	[CP[C [sigümjileV gsenT] gejüC]] bodu-na

This shows that inside the embedded clause in (53B‘), the verb moves to T and then the complex (the verb and the tense element) moves to C, which is followed by ellipsis of TP (indicated with grey shading). Sakamoto and Bao support this analysis with the possibility of adjunct-including readings and the impossibility of voice mismatch (see Sakamoto and Bao 2019 for details).

Assuming with Sakamoto and Bao (2019) that V-to-C movement indeed occurs in Mongolian, we may posit the following structure for (51a):

(55)	Bi-Ø [CP [TP Mergen-Ø [VP suruγči tV] tT] [C [bolV χuT] C]-yi (ni) mede-ne

In the complement clause, the copula verb moves to C via T, resulting in a complex complementizer indicated in boldface. Just like the empty complementizer in English, this complex complementizer may be omitted on the condition that the resulting empty category be lexically governed.¹⁵

The impossibility of copula drop in (51b) may be accounted for in the following way:

(56)	[CP [TP Mergen-Ø [VP suruγči tV] tT] [C [bolV χuT] C]-Ø ni nada-du mede-gde-ne

Here too, V-to-C movement via T takes place inside the subject clause. The complex complementizer may not be omitted, because it is not lexically governed just as the empty complementizer in (52b) is not. As for (51c), on which our informants’ reactions were split, it may be that for those who did not allow copula drop or complementizer omission, the lexical government condition applies to the surface representation, where the complement clause is dislocated from the complement position of VP, whereas those who tolerated it apply the condition after the scrambled clause is reconstructed. We leave it to our future research to elaborate on this line of analysis.

6 Conclusions and open issues

We have considered reduced embedded questions in CM and proposed a pseudo-sluicing analysis for them. We have shown that it can directly explain the lack of the case-matching effect, first observed for KM by Sakamoto (2012, 2015). According to our analysis, remnant wh-phrases are complements of the copula and hence are not assigned case, which is why they do not match their correlates in case. Our analysis is supported further by the fact that reduced questions can actually contain overt pronominal subjects, which is expected because null pronominal subjects posited by the pseudo-sluicing analysis should be able to alternate with their overt counterparts. We have also observed the fact that reduced questions in CM can be felicitous without linguistic antecedents, which reinforces our assumption that they do not involve ellipsis but pronominal subjects and the optional copula.

Before ending this paper, we would like to point out that our analysis does not entirely preclude the case-matching effect from emerging in CM. As observed by Sakamoto (2012) for KM, matrix sluicing in CM seems to be faithful to the case-matching effect (see Lasnik 1999 for some discussions on matrix sluicing). Consider the following data:

(57)	A:	Batu-Ø	nige	xümün-dü	ene	nom-i	ög-be.
		Batu-nom	one	person-dat	this	book-acc	give-pst
		‘Batu gave this book to a person.’
	B:	Xen-dü	bui?
		who-dat	prt
		‘To whom?’

Taking speaker A’s utterance as its antecedent, speaker B’s utterance consists of a wh-phrase and the question marker, though it can have the same interpretation as its full-fledged counterpart (namely, To whom did Batu give this book?). Notice that the remnant wh-phrase is marked dative just like its correlate in (57A).

In addition, when a reduced embedded question contains more than one remnant wh-phrase (namely, when we have a case of multiple sluicing (Takahashi 1994; Lasnik 2014; Abels and Dayal 2017), CM does exhibit the case-matching effect, as indicated below (Sakamoto (2012) observes comparable data in KM).

(58)	a.	Batu-Ø	nige	γaǰar-ača	nige	xümün-dü	beleg
		Batu-nom	one	place-abl	one	person-dat	present
		ilege-be,
		send-pst
		‘Batu sent a present to a person from a place,’
	b.	gebečü	bi-Ø	[tere-Ø	χamiγa-ača	xen-dü	beleg
		but	I-nom	he-nom	where-abl	who-dat	present
		ilege-gsen]-i	(ni)	mede-xü	ügei.
		send-perf-acc	ppc	know-inf	not
		‘but I don’t know to whom he sent a present from where.’
	c.	gebečü	bi-Ø	[χamiγa-ača	xen-dü]-yi	ni	mede-xü
		but	I-nom	where-abl	who-dat-acc	ppc	know-inf
		ügei.
		not
		‘lit. but I don’t know to whom from where.’

Anteceded by (58a), (58b-c) contain a full-fledged and a reduced embedded question, respectively. In (58c), the two wh-phrases are case-marked in the same way as their correlates in (58a). Close considerations of cases like (57) and (58) are left to future research.