Morphosyntactic Marking of Focus: Subject–Object Asymmetries in Bantu

Abstract

In many African languages, there exists a type of subject–object asymmetry by which subject focus must be expressed by A-bar movement to a morphologically marked left peripheral position whereas object focalisation can be expressed by movement and morphological marking in the left periphery, or in situ. In this article, we discuss and analyse this structural asymmetry in the Bantu languages Basaá and Mmaala. We argue that overt and covert movement of the focused object to Spec-FocP in the left periphery is allowed while covert movement of the subject is blocked, so that overt movement is the only possible option. Contrary to previous analyses, which attribute the obligatoriness of subject focus movement and marking to an interpretive conflict, we propose a formal characterisation of this phenomenon by which the blocking of subject focalisation in situ is deduced from criterial freezing, so that overt movement to the left periphery is the only option, through a familiar strategy of overt subject extraction.

1. Introduction1

The question as to why subject focus cannot be expressed in situ in the canonical subject position in many African languages has raised much attention from various viewpoints in the literature. The prevailing claim is that preverbal subjects in those languages have topic-like properties (see Bresnan & Kanerva, 1989; Morimoto, 2000; Henderson, 2006; Zeller, 2008; on Southern Bantu and Fiedler et al., 2010 on Chadic, Kru and Kwa), in line with Givón (1976). According to this view, the subject must be ‘detopicalised’ in order to allow a narrow focal interpretation. This is generally obtained in two ways: Either the subject is placed postverbally, or it is moved to the left periphery. Alternative views have been put forth on the exact nature of the interpretive properties of subjects (e.g., Zerbian, 2006; Cheng & Downing, 2014), which share with the previously mentioned approaches the idea that the obligatory marking of subject focus is fundamentally due to an interpretive clash that would otherwise arise. In short, all these claims try to deduce the subject/object asymmetry from interpretative effects. In this paper, we explore a type of subject–object asymmetry that occurs in focus constructions in the Bantu languages Basaá and Mmaala. Like in many other Bantu languages, subject focus involves obligatory movement in overt syntax to a dedicated position in the left periphery, a position marked by special focus morphology, whereas object focus can be expressed by overt movement with morphological marking, or in situ. We show that this asymmetry in focalisation is not attributed to an interpretative conflict but is a consequence of criterial freezing (Rizzi, 2006, 2021). The paper is organised as follows: Section 1 provides background information about Basáa and Mmaala. Section 2 is concerned with subject and object focus marking. Section 3 deals with the syntactic analysis of the phenomena under discussion.

2. Background

Basaá (ɓasaá) and Mmaala are tone2 languages that belong to the Narrow Bantu group. They are spoken in Cameroon and display a basic SVO order like in (1) and (2).

(1)	málêt	a-n-sɔ́mb	βíkaat	Basaá
	1teacher	1sm-pst1-buy	8books
	‘The teacher bought the books.’

(2)	agánd	gɔ-gu-ɔn	iɲɔk	Mmaala
	cl.woman	fut1-inf-kill	snake
	‘The woman will kill a snake.’

In double object constructions, the indirect object precedes the direct object, and the reverse order is disallowed.

(3)	málêt	a-n-tí	(*bikaat)	ɓáúdú	βikaat	Basaá
	1teacher	1sm-pst1-give	8books	2students	8books
	‘The teacher gave the books to the students.’

(4)	agánd	gɔ-gu-fa	(*maɲan)	baabajɔ	maɲan	Mmaala
	cl.woman	fut1-inf-give	food	cl.children	cl.food
	‘The woman will give the food to the children.’

In both languages, the direct object > indirect object order is allowed if the former is pronominalised. This occurs when the direct object represents old information whereas the indirect object is new in the discourse. This is illustrated in the question–answer pairs in (5) and (6).

(5)	A.	ndӡέɛ́	ɓaúdú	ɓá-n-tí	mákebla	Basaá
		1who	1teacher	2sm-pst1-give	6presents
		‘Who did the students give the presents to?’
	B.	ɓaúdú	ɓá-n-tí	mɔ́	malêt	(*mɔ́)
		1teacher	1sm-pst1-give	6them	1teacher	6them
		‘The students gave them to the teacher.’

(6)	A.	aɲa	agánd	gɔ-gu-nu-fa	maɲjan	Mmaala
		who	woman	fut1-inf-prt-give	food
		‘Who will the woman give the food to?’
	B.	agánd	gɔ-gu-nu-fa	mɔ́	baabajɔ	(*mɔ́)
		cl.woman	fut1-inf-give	cl.it	cl.children	cl.it
		‘The woman will give it to the children.’

Basaá and Mmaala are noun class languages. As is the case in many languages, nouns are classified according to classes. Noun classes are indicated by specific prefixes. Basaá involves twelve distinct nominal classes, which vary in number and person. At this point of the study, we are not aware of any work on the noun class system on Mmaala. We refrain from using cardinal numbers for the glossing of the noun classes in this language. Based on its rich morphological agreement displayed in the nominal system common to Bantu languages, we assume that we are dealing with a noun class language akin to many other noun class languages. The following nominal constructions in (7) and (8) delineate morphological agreement between the head noun and its dependents.

(7)	a.	íní	tâmb	jɛ̂m	ikέŋí	ilâm	Basaá
		7dem.prox	7shoe	7poss.3pl	7big	7nice
		‘This nice (as opposed to that) big shoe of mine.’
	b.	βíní	βitámb	gwɛ̂m	βikέŋí	βilâm
		8dem.prox	8shoes	8poss.1sg	8big	8nice
		‘These (as opposed to those) nice big shoes of mine.’

(8)	a.	gibɔ́gɔd	gjam	gibólotot	egje,	Mmaala
		cl.shoe	cl.poss.1sg	cl.red	dem.prox.cl
		‘This red shoe of mine.’
	b.	bibɔ́gɔ̀d	bjam	bitatɔ	bibólotot	ebje
		cl.shoes	cl.poss.1sg	cl.three	cl.red	dem.prox.cl
		‘These three red shoes of mine.’

Morphological agreement in the noun phrase is controlled by the head noun, the morphological features of which are shared by its dependents. Basaá and Mmaala also differ with respect to the structural position of negation and subject–verb agreement.

(9)	a.	málêt	a-bí-tí	ɓê	ɓáúdú	bikaat	Basaá
		1teacher	1sm-pst2-give	neg	2students	8books
		‘The teacher did not give the books to the students.’
	b.	agánd	di-ga-gu-fa	baabajɔ	maɲjan	Mmaala
		cl.woman	neg-fut1-inf-give	cl.children	cl.food
		‘The woman will not give the food to the children.’

Basaá (9a) has low negation that occurs after the verb whereas negation is higher in Mmaala (9b). Negation precedes the verb and all the elements that make up the verbal complex. Another structural difference between both languages is the null subject phenomenon (Rizzi, 1982). Despite the fact that Mmaala displays morphological agreement within the noun phrase (8), it is characterised by the absence of morphological subject–verb agreement and does not allow null subjects (11). Unlike Mmaala, Basaá (10) displays morphological subject–verb agreement and allows null subjects like many other Bantu languages.3

(10)	a.	málêt	a-bí-tí	ɓáúdú	bikaat
		1teacher	1sm-pst2-give	2students	8books
		‘The teacher gave the students the books.’
	b.	a-bí-tí	ɓɔ́	bí-kaat
		1sm-pst2-give	2them	8-books
		‘S/he (the teacher) gave them the books.’

(11)	a.	agánd	ga-gu-fa	baabajɔ	maɲan
		cl.woman	fut1-inf-give	cl.children	cl.food
		‘The woman will give the children the food.’
	b.	*ga-gu-fa	bɔ́	maɲan
		fut-inf-give	2them	cl.food

Just like in other languages that delineate morphological subject–verb agreement, null subjects in Basaá are licensed by the subject marker (sm) that bears the noun class features of the subject, also encoding person and number specifications. The presence of morphological agreement inside the nominal layer and its absence at the clausal level in Mmaala make it different from Basaá and other noun class Bantu languages such as Tuki (Biloa, 2013), Kinande (Baker, 2008), Makhuwa-Enahara (Van der Wal, 2009), Sambaa (Riedel, 2009), and Zulu (Zeller, 2008; Halpert, 2015) in which morphological agreement occurs both within the nominal and clausal layers, supporting the view that DP-internal agreement and subject–verb agreement are different morphosyntactic phenomena involving distinct mechanisms (see Moscati & Rizzi, 2014, based on acquisition data). In comparison with Mmaala, Basaá is relatively well described in the literature. Some works include grammatical studies such as Rosenhuber (1908); Schürle (1912); Voorhoeve (1980); Dimmendaal (1988); Bitjaa Kody (2017), etc. Some well-known works in linguistics include Bôt Ba Njock (1962, 1970), Bitjaa Kody (1990), Mbom (1990), Schmidt (1994, 1996), Ngo Ndjeyiha (1996, 2005), Mutaka and Bitjaa Kody (2000), Hyman (2003), Makasso (2008, 2023), Bassong (2014, 2019, 2021), Hamlaoui and Makasso (2015), Makasso and Lee (2015), etc. Coded A.62B by Maho (2009), the Mmaala language has been studied by Ossoko (2017, 2019, 2023) on pragmatics.

2.1. Focus Marking

This section discusses subject and non-subject focalisation in the two languages. Like in many African languages, there is a salient structural asymmetry between subject and non-subject focus marking. See also (F. Schwarz, 2003, 2007; Bassong, 2014 for other Bantu, Zimmermann, 2006; Hartmann & Zimmermann, 2007 for A. Schwarz & Fiedler, 2007; Fiedler et al., 2010 for Chadic, Kwa and Gur). We confine ourselves to subject and object focalisation as focalisation of adjuncts, as well as of indirect objects, is analogous to that of objects. While subject focus involves overt movement to the left periphery, to a position signalled by a dedicated focus marker in Basaá and Mmaala, object focus admits both movement to the left periphery and the in situ strategy, by which the focused constituent occurs postverbally and is not morphologically marked in both languages.

2.2. Subject Focus Marking

Although subject focalisation has to always occur ex situ, each language uses a specific focus marking strategy. Subject focus can encode new and contrastive information as discussed in the following sections.

2.2.1. Information Focus

Starting from the basic order in (12), and using the question–answer congruence as a test for focushood, it can be observed in (13)B that the focalised subject malêt ‘teacher’ in Basaá needs to be moved and focus marked. As is generally the case when focus is expressed by movement, the subject is followed by the focus marker -n, which adjoins to the corresponding agreement morpheme ɲέ- giving rise to the composite form ɲέ-n. More precisely, the morpheme ɲέ- bears the gender (noun class 1) and number (singular) specification of the subject. The unacceptability of (13)B indicates that a focalised subject cannot be left in situ, hence the simple canonical SVO order is infelicitous4 as a reply to a subject wh-question that requires narrow focus on the subject.

(12)	malêt	a-n-tíl-á	↓βíkaat	Basaá
	1teacher	1sm-pst1-write-fv	8.books
	‘The teacher wrote the books.’

(13)	A.	ndӡέέ	a-n-tíl-á	↓βíkaat	Basaá
		1who	1sm-pst1-write-fv	8.books
		‘Who wrote the books?’
	B.	malêti	ɲέ-n	ti	a-n-tíl-á	↓βíkaat
		1teacher	1-foc		1sm-pst1-write-fv	8.books
		‘The teacher wrote the books.’
	C.	#malêt	a-n-tíl-á	↓βíkaat
		1teacher	1sm-pst1-write-fv	8.books
		‘The teacher wrote the books.’

Unlike in languages like Gungbe (Aboh, 2004), Kikuyu (F. Schwarz, 2007) and Tuki (Biloa, 2013), question words (wh-phrases) in Basaá cannot occur with overt focus morphology (13)A, whereas the corresponding focused constituent in the answer does (13)B. Clearly, a low-level parameter is involved here (does wh- co-occur with an overt focus marker? Tuki: yes; Basaa: no).

Compared to Basaá, Mmaala does not have morphological focus markers. Nevertheless, the focusing of a subject in this language is expressed

A.

By the use of a resumptive pronoun in subject position;

B.

By a morphological change in the verbal complex (addition of a particle, a reflex of A-bar extraction);

C.

By the optional addition of a cleft-like structure (it is...) at the beginning of the clause

These three changes are found both in the question (15)A and in the answer (15)B, which are derived from the basic order in (14).

(14)	endimen	am	gɔ́-gú-ɔ̂n	iɲɔk	Mmaala
	brother	1sg.poss	fut1-inf-kill	cl.snake
	‘My brother will kill a snake.’

(15)	A.	(bu		ɲɛ)		aɲà	ui		Mmaala
		expl		be		who	3sg
		gɔ́-gú-nú-ɔ̂n		iɲɔk
		fut1-inf-prt-kill		cl.snake
		‘Who will kill a snake?’
	B.	(bu	ɲɛ)			endimeni	am		Mmaala
		expl	be			brother	1sg.poss
		ui	gɔ́-gú-nú-ɔ̂n			ɲέ  /	iɲɔk
		3.sg	fut1-inf-prt-kill			3sg	cl.snake
		‘It is my brother who will kill a snake/my brother will kill it (a snake).’
	C.	#endimen	am		gɔ́-gú-ɔ̂n		ɲέ  /	iɲɔk	Mmaala
		brother	1sg.poss		fut1-inf-kill		3sg	cl.snake
		‘My brother will kill it (a snake).’

The crucial point that Mmaala has in common with Basaá is that narrow focus on the subject cannot be expressed in situ, i.e., (15)C is not a felicitous answer to (15)A.

So, the obligatoriness of movement to express subject focus is also observed in Mmaala.

A few comments are required here on certain morphosyntactic peculiarities of the language. Mmaala may use a full-fledged cleft structure in this context, or simple movement to the left periphery (the optionality of the cleft structure is indicated by the parentheses where both the expletive bu ‘it’ and the copula ɲɛ ‘be’ can be omitted). Whether or not a full cleft structure is used, as shown in (15)A and (15)B, subject questions and subject focus are obligatorily marked through 1. a resumptive pronoun in subject position, and 2. a morphological change in the verbal complex. In (15), the subject wh-phrase aɲà ‘who’ and the focalised subject endimen am ‘my brother’ are resumed by the third person resumptive pronoun u ‘he’. The resumptive strategy is accompanied by the appearance of the particle nú sandwiched between the future tense morpheme gɔ́ and the infinitive particle gú: a system of such particles marks A’ movement from the IP in this language, a property reminiscent of “wh-agreement”5 phenomena in many other languages, which we intend to analyse in future work. Like in the Basaá example in (13)C, the canonical SVO order in Mmaala (15C) is infelicitous in this question–answer context, hence, focalisation in situ of subjects is not possible. Subject focus can also be used to encode a contrastive or corrective interpretation using the same morphosyntactic devices, as discussed in the next section.

2.2.2. Contrastive/Corrective Focus

In the presence of a set of salient or implied alternatives in the discourse (Rooth, 1992, 1996), subject focus can be used to encode contrast or correction in both languages. We consider the context in (16)6 to highlight corrective focus in Basaá whereby the focalised subject ɓaúdú ‘students’ in (16)B contrasts with the coordinate subject ɓagwâl ni ɓaúdú ‘parents and students’ in (16)A. The focalised subject ɓaúdú ‘students’ is followed by the agreeing focus marker ɓɔ́-n.

(16)	A.	ɓaá	ɓagwâl	ni	ɓaúdú	ɓá-bí-kɛ	í	ɓɔmá	a Basaá
		qm	2parents	and	students	2sm-pst2-go	loc	7meeting	qm
		‘Did the parents and students go to the meeting?’
		Tɔ,	ɓaúdú	ɓɔ́-n ti	ɓá-bí-kɛ	(í	ɓɔmá)
	B.	no	2parents	2-foc	2sm-pst2-go	loc	7meeting
		‘No. The parents (not the parents and the students) went (to the meeting).’

Similarly, in the micro discourse in (17), the focalised subject mɔɔnd ‘man’ in Mmaala (17)B stands in contrast with ɔmbúlɔ ‘girl’ in (17)A. In (17)B focalisation of the lexical subject endimen am ‘my brother’ is signalled by the presence of the third person resumptive pronoun u ‘he’ and the particle nú in the verbal complex.

(17)	A.	iŋ		gú-bidaga		kaa		ɔmbúlɔ		Mmaala
		1sg		prs-think		that		girl
		gɔ́-gú-ɔ̂n		iɲɔk
		fut1-inf-kill		cl.snake
		‘I think that a girl will kill a snake.’
	B.	ǎko	(bu		ɲɛ)		endimen		ami
		no	expl		be		Brother		1sg.poss
		ui	gɔ́-gú-nú-ɔ̂n		ɲέ  /		iɲɔk
		3.sg	fut1-inf-prt-kill		3sg		cl.snake
		‘No. It is my brother who will kill it/a snake/No. my brother (not a girl) will
		kill it/a snake.’

In (16) and (17), contrastive/corrective focus is expressed by the mention of explicit sets of comparable alternatives in the discourse. Summarising, the earmark of subject focalisation, be it in new information or contrastive contexts, is that it always involves movement and is marked morpho-syntactically in different ways in the two languages. The focalised subject needs to occur ex situ and by the use of a focus marker in Basaá. In Mmaala, it is signalled by subject resumption and the morphological change in the verbal complex signalling extractions. This brings up the question as to why subject focus marking is obligatory in Basaá and Mmaala like in many other African languages. The answer to this question is given to Section 3.

2.3. Object Focus Marking

Unlike subject focalisation that has to be marked morphosyntactically, object focus can be expressed in situ without any grammatical encoding. This is possible in information and contrastive focus situations in both languages. Examples (18) and (19) illustrate both situations in Basaá while (20) and (21) are instances of new information and contrastive focus in Mmaala.

(18)	A.	Kíí	mudaá	a-ń-lâmb	Basaá
		10what	1woman	1sm-prs-cook
		‘What is the woman cooking?’
	B.	(mudaá)	a-ń-lâmb	tꭍɔ́βī
		1woman	1sm-prs-cook	13fish
		‘(The woman) she is cooking fish.’

(19)	A.	í	ḿ-pooná	lέ	mudaá	a-n-lâmb	kóp	Basaá
		expl	prs-seem	that	1woman	1sm-pst1-cook	9chicken
		‘It seems that the woman cooked chicken.’
	B.	hέέní.	mudaá	a-n-lâmb	tꭍɔ́βī
		no	1woman	1sm-pst1-cook	13fish
		‘No. (The woman) she cooked fish (not chicken).’

(20)	A.	giàg	Ambassa	gɔ́-gú-nú-ɔ̂n	Mmaala
		what	Ambassa	fut1-inf-prt-kill
		‘What will Ambassa kill?’
	B.	Ambassa  /	u	gɔ́-gú-ɔ̂n	iɲɔk
		Ambassa	3sg	fut1-inf-kill	snake
		‘Ambassa/he will kill a snake.’

(21)	A.	Ambassa	ɔ	gu-ɔ̂n	mbwa	á	Mmaala
		Ambassa	cj	pst3-kill	dog	qm
		‘Did Ambassa killed a dog?’
	B.	(ǎkó).	Ambassa  /	u	ɔ	gu-ɔ̂n	iɲɔk
		no	Ambassa	3sg	cj	pst3-inf-kill	snake
		‘No. Ambassa/he killed a snake (not a dog)?’

Object focus can also be expressed by fronting the object to the left periphery. Like subject focus, left peripheral focalisation of the object in Basaá requires morphological focus marking on the object. This strategy is felicitous in information and contrastive focus contexts as shown in (22) and (23), respectively. The fronted objects agree in noun class with the focus marker.

(22)	A.	Kíí	maaŋgέ	a-ń-dӡowa	Basaá
		9what	1child	1sm-prs-wash
		‘What is the child washing/What does the child wash?’
	B.	mambɔ́t	mɔ́-n	(maaŋgέ)	a-n-dӡowa
		6clothes	6-foc	1child	1sm-prs-wash
		‘The child/he is washing the clothes.’

(23)	A.	maaŋgέ	a-ń-dӡowa	mambɔ́t	ɛ	Basaá
		1child	1sm-prs-wash	6clothes	qm
		‘The child is washing/washes the clothes.’
	B.	tɔ.	bitámb	gwɔ́-n	(maaŋgέ)	a-ń-dӡowa
		no	8shoes	8-foc	1child	1sm-prs-wash
		‘No. The child is washing the shoes.’

Like subject focalisation, object focalisation through movement to the left periphery may or may not involve a cleft structure and triggers a morphological change in the verbal complex in Mmaala. However, unlike subject focalisation, ex situ focalisation of the object does not trigger resumption at the base as shown in (24) and (25). The particles ɲá in (24) and nú (25)B are reflexes of A-bar movement in the “past tense three” form (“yesterday’s past”) and the future tense, respectively.

(24)	A.	giàg	Ambassa	ɲá-gu-ɔ̂n	Mmaala
		what	Ambassa	prt-pst3-kill
		‘What did Ambassa kill?’
	B.	(bu	ɲɛ)	iɲɔk	Ambassa /	u	ɲá-gu-ɔ̂n	(*jέ)
		expl	be	cl.snake	Ambassa	3sg	prt-pst3-kill	it
		‘Ambassa (he) killed a snake.’

(25)	A.	iŋ	gú-bidaga	kaa	Ambassa		Mmaala
		1sg	prs-think	that	Ambassa
		gɔ́-gú-ɔ̂n	mbwa
		fut-inf-kill	dog
		‘I think that Ambassa (he) killed a dog.’
	B.	(ǎkó)	(bu	ɲɛ)		iɲɔk
		no	expl	be		cl.snake
		Ambassa  /	u	gɔ́-gú-nú-ɔ̂n		(*jέ)
		Ambassa	3sg	fut1-inf-prt-kill		it
		‘(No), Ambassa (he) will kill a snake (not a dog).’

In comparison with (13)–(15), the preceding examples in (18)–(24) show the structural asymmetry between the subject and object with respect to focus realisation. This subject–object asymmetry observed cross-linguistically in many African languages has not yet received a consensual analysis. There are divergent views with respect to the main motivations for the obligatoriness of subject focus marking in many languages. Works such as Bresnan and Kanerva (1989), Morimoto (2000), Henderson (2006), and Zeller (2008) on Southern Bantu languages and Fiedler et al. (2010) on Chadic, Kru and Kwa languages suggest that subject focus marking simply follows from Givón’s (1976) generalisation that preverbal subjects have topic-like properties, an idea which is also in line with Li and Thompson (1976)’s view that subjects and topics share some common properties. The aboutness property (Li & Thompson, 1976; Rizzi, 2005; Belletti & Rizzi, 2017), one of such salient features, has been used by some Africanists as evidence that subject focus marking is a detopicalisation strategy. In other words, according to this approach, the position of the lexical subject in the SVO order is topical. In order to achieve a narrow focal interpretation, the subject must be “detopicalised” through overt focus marking.

This analysis has the appeal of simplicity. Nevertheless, we believe there is evidence against it. Whereas there is a tendency to associate subjecthood and topicality, it is not the case that the subject position is necessarily interpreted as topical. For instance, indefinite elements like someone, something, etc. (see Rizzi, 1997; Frascarelli, 2000; Belletti & Rizzi, 2017 for Italian; Zerbian, 2006 for Nothern Sotho; Van der Wal, 2009 for Makhuwa-Enahara; Halpert, 2012 for Zulu; Cheng & Downing, 2014 for Zulu) are not natural topics, but can occur in the preverbal subject position. This can be observed in Basaá (26) and Mmaala (27), where the indefinite quantifiers ŋgim mut and mɔɔnd, the English equivalents for ‘someone’, function as lexical subjects.

(26)	ŋgim	mut	i-bí-sómb-ôl	ɓɔ́ɔ́ŋgέ	bitámb	Basaá
	1indf	1person	9sm-pst2-buy-appl	2children	shoes
	‘Someone bought the shoes for the children.’

(27)	mɔɔnd	sa-fa	babàjɔ	màɲán	Mmaala
	someone	pst2-give	children	food
	‘Someone gave the food to the children.’

Moreover, inherently non-topical elements like ŋgim mut and mɔɔnd ‘someone’, if focalised, require overt movement and cannot be focalised ‘in situ’ in the canonical subject position. This is shown in (28) for Basaá and (29) for Mmaala. The absence of subject movement in subject focus contexts yields an infelicitous interpretation as indicated in the C’s replies.

(28)

A.

ndӡέέ

a-bí-sómb-ôl

ɓɔ́ɔ́ŋgέ

bitámb

Basaá

1who

1sm-pst2-buy-appl

2children

shoes

‘Who bought the shoes for the children.’

B.

ŋgim

mut

jɔ̌-n

i-bí-sómb-ôl

ɓɔ́

gwɔ̄

1indf

1person

9-foc

9sm-pst2-buy-appl

2them

8them

‘Someone bought them for them.’

C.

#ŋgim

mut

i-bí-sómb-ôl

ɓɔ́

gwɔ̄

1indf

1person

9sm-pst2-buy-appl

2them

8them

‘Someone bought them for them.’

(29)

A.

(bu

ɲɛ)

aɲai

ui

sa-gu-fa

babàjɔj

màɲánk

Mmaala

expl

be

who

3sg

pst2-prt-give

children

food

‘Who gave food to the children?’

B.

(bu

ɲɛ)

mɔɔndi

ui

sa-gu-fa

bɔ́j

mɔk

expl

be

person

3sg

pst2-prt-give

3pl

3pl

‘someone gave it to them.’

C.

#mɔɔnd

sa-fa

bɔ́j

mɔk

person

pst2-1sg-give

3pl

3sg

‘Someone gave it to me.’2

Examples (28)B and (29)B attest that an interpretive conflict approach to (13)C and (15)D is not adequate as existential quantifiers that are not natural topics also have to be focus marked ex situ. Thus, the subject position in the canonical SVO order cannot be assumed to obligatorily trigger topic information. Put another way, the obligatoriness of subject focus movement and marking cannot be analysed as an interpretative conflict resolution strategy.

3. Criterial Freezing and the Subject-Criterion

In this section, we show how the obligatory ex situ focalisation of subjects can be captured in formal terms. More precisely, we show that criterial freezing is the main source of this phenomenon in these languages, and probably in many other African languages from various groups such as Bantu (see Zerbian, 2006 for Nothern Sotho; Van der Wal, 2009 for Makhuwa-Enahara; Halpert, 2012 for Zulu; Cheng & Downing, 2014 for Zulu, etc.), Chadic, Kwa and Gur (see Fiedler et al., 2010 for a cross-linguistic study, but see also Hartmann & Zimmermann, 2007 for Hausa), which exhibit similar subject–object asymmetries in focalisation. In line with the criterial approach to scope discourse semantics, chain formation in syntax (Rizzi, 2006, 2017; Rizzi & Shlonsky, 2007; Shlonsky, 2014; Shlonsky & Rizzi, 2018; Roy & Shlonsky, 2019, etc.) is sensitive to structural configurations so much so that phrasal categories moved to the specifier positions of criterial heads are frozen in place and become unavailable to further movement, along the lines of the principle in (30).

(30)	Criterial Freezing: In a criterial configuration, the criterial goal is frozen in place.
	(Rizzi, 2021, p. 275).

Starting from Rizzi (2006), the principle in (30) guarantees the mapping between a phrasal category, a given syntactic configuration and a specific interpretative property. As such, the driving force of a criterial head is that once a given phrase with matching features is attracted to its specifier position, that phrase gets frozen and becomes unavailable for further syntactic operations.7 At the interfaces with sound and meaning, this criterial head guides the interpretative properties of its specifier and complement and yields configurations such as Topic–Comment, Focus–Presupposition, etc. Criterial freezing is not only concerned with phrasal categories moving to scope taking positions in the left periphery. It also deals with phrases targeting A-positions such as the subject position. In this section, the analysis is confined to subject and focus positions. Rizzi (2006) suggests that the subject position is criterial following the cartography of subject positions (Cardinaletti, 2004). According to him, there is a formal Subj(ect) head in the clausal spine. This Subj head is merged above the T head. It attracts the closest nominal element to its Spec and, at LF, it determines the “aboutness” property: the complement of Subj, the predicate, is going to be interpreted as being “about” the specifier. Being a criterial position, when the lexical subject is moved to Spec-SubjP it undergoes freezing effects as a consequence of (30).

Consider now the impossibility of focalisation in situ of the subject, illustrated by the C answers in (31) and (32).

(31)	A.	ndӡέέ	a-bí-tī-	ɓɔ́ɔ́ŋgέ	bikaat	Basaá
		1who	1sm-pst2-give	2children	8books
		‘Who gave the books to the children?’
	B.	malêt	ɲέ-n	a-bí-tī-	ɓɔ́	gwɔ̄
		1teacher	1-foc	1sm-pst2-give	2them	8them
		‘The teacher gave them to them.’
	C.	#malêt	a-bí-tī-	ɓɔ́	gwɔ̄
		1teacher	1sm-pst2-give	2them	8them
		‘The teacher gave them to them.’

(32)

A.

(bu

ɲɛ)

aɲai

ui

sa-gu-fa

babàjɔj

màɲánk

Mmaala

expl

be

who

3sg

pst2-prt-give

children

food

‘Who gave food to the children?’

B.

(bu

ɲɛ)

agándi

ui

sa-gu-fa

bɔ́j

mɔk

expl

be

woman

3sg

pst2-prt-give

3pl

3pl

‘The woman gave it to them.’

C.

#agánd

sa-fa

bɔ́j

mɔk

woman

pst2-1sg-give

3pl

3sg

‘The woman gave it to me.’

Assuming that ‘in situ’ focalisation involves covert focus movement (Chomsky, 1977), and that covert as much as overt focus movement targets a dedicated Spec-FocP in the left periphery (Rizzi, 1997), this covert movement cannot occur from a subject position, which, following the literature quoted above, we assume to be criterial. Thus, the unacceptability of (31)C and (32)C derived in (33a) and (33b) arises from the fact that covert subject movement to Spec-FocP is blocked by criterial freezing in Spec-SubjP as depicted in (33).

(33)

Why is it that, if covert subject movement to focus is excluded by freezing, overt subject movement is fine? It is well-known that natural languages can use a variety of strategies to overtly move subjects and circumvent freezing effects (Rizzi & Shlonsky, 2007). One classical strategy generally available to Null Subject Languages consists of satisfying the Subject Criterion (or, in GB terms, the Extended Projection Principle) through non-referential pro, and moving the thematic subject from a lower position (Rizzi, 1982). As Basaá is a null subject language, we expect this strategy to be available. So, in (34a) the subject-criterion may be satisfied by pro in Spec-SubjP, and the thematic subject may be moved directly from the thematic position to the Spec-Foc position.8

On the contrary, Mmaala is not a Null Subject Language: it lacks morphological subject–verb agreement, hence, in accordance with Taraldsen’s generalisation, it does not license subject pro. All other things being equal between the two languages, we expect Mmaala to permit overt subject focalisation but with the use of an overt resumptive pronoun, pro being unavailable. The expectation is correct, as is shown by the acceptability of (32b). Under the set of assumptions we are adopting, this sentence is derived as indicated in (34b), with Spec-SubjP filled by an overt pronoun.9 The lexical subject in both languages takes off directly from Spec-vP, its first merge site on its way to Spec-FocP (or through the alternative derivational steps mentioned in fn 7). In Basaá (34a), the focus head is filled by the focus marker ɲέ-n, which values its features with those of the focalised subject malêt ‘teacher’ in Spec-FocP. Again, it is assumed that Mmaala uses an unpronounced focus marker akin to Italian-type languages (Rizzi, 1997) whose focus markers are implicit.

(34)

This analysis straightforwardly captures the subject–object asymmetry. In situ focalised objects as described in (18)–(21) are allowed because there is no object-criterion, hence no freezing effect. More precisely, in cases like (18)–(21), covert object focus movement on the LF side is allowed because nothing blocks the object from moving.10 Alongside in situ focalisation (hence covert movement to FocP), overt object movement to focus is an alternative option: both are consistent with the operative principles and lead to well-formedness. In partial conclusion, the obligatoriness of subject focus marking in Basaá and Mmaala can be traced back to a formal explanation, namely criterial freezing. Freezing effects block covert movement of the lexical subject to Spec-FocP in canonical SVO structures. To circumvent these effects and satisfy both the subject and focus criteria, each criterion has to be satisfied independently. The subject-criterion is satisfied by a resumptive subject pronoun in Spec-SubjP, null or overt depending on whether a language allows null subjects or not. In so doing, the subject moves directly from its Spec-vP to the focus field (or is remnant-moved, in the derivation suggested in fn 7) where it can satisfy the focus-criterion successfully.

4. Conclusions

The structural asymmetry observed in subject and object focalisation in the Bantu languages Basaá and Mmaala is also attested in many other African languages. In this paper we have argued against an analysis based on an interpretive conflict between subjects and topics. We have proposed an alternative by which the obligatoriness of subject focus marking is deduced from Criterial Freezing. Covert movement from a subject position is excluded as a freezing effect, whereas covert object movement is unimpeded, as there is no object criterion. Overt movement to Spec-Foc of a subject is made possible by a familiar rescue strategy, the satisfaction of the Subject Criterion by a pronoun, a strategy immediately visible in the non-Null Subject Language Mmaala. This class of subject–object asymmetries in focalisation strategy is thus reduced to the larger family of the subject–object asymmetries of the that-trace type.