Focus Fronting in a Language with In Situ Marking: The Case of Mǝ̀dʉ́mbà

Abstract

This paper discusses the structural realisation of contrastive focus in the Grassfields Bantu language Bamileke Mǝ̀dʉ́mbà, yet another language with grammatically optional focus fronting. We show that the realisation of contrastive focus in Mǝ̀dʉ́mbà is by default marked in situ with the morphological focus marker á. We further show that á introduces an additional, not-at-issue exhaustivity inference as part of its lexical meaning. Taken together, this means that two often-discussed interpretive triggers of grammatically optional focus fronting, namely, contrastivity and exhaustivity, are not responsible for triggering Mǝ̀dʉ́mbà’s focus left-dislocation. Rather, the main semantic contribution of focus fronting consists in triggering an existence presupposition, a discourse-semantic effect well known from other focus–background bipartitions, possibly in combination with other, softer discourse-semantic effects, such as the special emphasis required in cases of discourse unexpectedness.

1. Introduction

Following up on a previous treatment by Keupdjio (2020), this article investigates the structural realisation of focus and the reasons for focus fronting in the Cameroonian Grassfields Bantu language Mǝ̀dʉ́mbà. We will show that Mǝ̀dʉ́mbà consistently realises contrastive focus in situ, where the notion of contrastive focus is understood in the sense of Rooth (1992), Krifka (2008), and Kratzer and Selkirk (2020), as invoking the presence of a set of contextually salient alternatives. In situ focus in Mǝ̀dʉ́mbà is marked by the morphological focus marker á, which immediately precedes the contrastive focus constituent. This is shown in (2a,b) for subject and direct object focus, respectively, as compared to the basic S-TAM-VOX sentence in (1) without explicit focus marking.1

(1)	nnά	náʔ	fά	tʃàŋ	nǔm	nmí
	Nana	pst.rem	give	food	to	Nami
	‘Nana gave food to Nami’.

(2)	a.	á	nyú	fǝ̀	ntʃɔ̀g	ὰm		[SUBJ focus]
		foc	snake	pst.near	bite	me
		‘A SNAKE bit me (… and not a scorpion)’.
	b.	nnά	náʔ	lùʔ	á	bí		[OBJ focus]
		Nana	pst.rem	take	foc	knife
		‘Nana took a KNIFE (…and not a pen)’.

With the verb focus in (3), focus marking additionally requires predicate doubling, the same as in many other West African languages (Zimmermann 2016). In this case, the finite verb occurs in its normal position following the TAM markers in regular S-TAM-VO order, but it cannot be focus-marked with the á marker. Instead, the á marker must precede a copy of the verb, which follows the main verb and its direct object, if present. Here, the term verbal copy is used in a pre-theoretical sense and refers to the doubled occurrence of V in a non-canonical position. Placing the á marker in front of a non-doubled finite verb results in ungrammaticality, cf. (3b).

(3)	a.	nnά	náʔ	lùʔ	bí	á	lùʔù	[V focus]
		Nana	pst.rem	take	knife	foc	take
		‘Nana TOOK the knife (…she did not steal it)’.
	b.	* nnά	náʔ	á	lùʔ	bí
		Nana	pst.rem	foc	take	knife

Also, observe that explicit focus marking with á is optional with in situ wh- questions and their corresponding answers. That is, á is not required for reasons of grammatical well-formedness. Wh- questions and their answers may simply request and provide new information in the absence of contextually salient alternatives (Krifka 2008; Kratzer and Selkirk 2020), cf. (4a). In contrast, á is required with D-linked wh- questions, which make reference to a contextually salient set of alternatives, cf. (4b). In such cases, the focus marker á must also occur in the corresponding answer. Note that the in situ strategy is the default way of forming wh- questions in Mǝ̀dʉ́mbà.

(4)

a.

Q:

nnά

náʔ

fά

tʃàŋ

nǔm

wʉ́

(new information: no á-marking!)

Nana

pst.rem

give

food

to

whom

‘To whom did Nana give food?’

A:

nnά

náʔ

fά

tʃàŋ

nǔm

nmí

Nana

pst.rem

give

food

to

Nami

‘Nana gave food to NAMI’.

b.

Q:

nnά

náʔ

fά

tʃàŋ

á

nǔm

wʉ́

(contrastive focus: á-marking!)

Nana

pst.rem

give

food

foc

to

who

‘To whom/which of them did Nana give food?’

A:

nnά

náʔ

fά

tʃàŋ

#(á)

nǔm

nmí

Nana

pst.rem

give

food

foc

to

Nami

‘Nana gave food to NAMI (and not to NUMI)’.

Finally, Keupdjio (2020, pp. 17, 75) showed that á-marked focus constituents can also occur fronted in the sentence-initial position, at least with individual denoting NPs; (5a,b) shows a minimal pair with the á-marked direct object in situ and ex situ (original diacritics and glossing adapted from Keupdjio 2020).

(5)	a.	Wàtɛ́t	náʔ	swɛ́ɛ̀n	á	Nùŋgɛ̀		(Keupdjio 2020, p. 17, ex. 4a)
		Watat	pst.rem	sell	foc	Nuga
		‘Watat betrayed NUGA’.
	b.	á	Nùŋgɛ̀	Wàtɛ́t	náʔ	swɛ́ɛ̀n	lá	(Keupdjio 2020, p. 18, ex. 4b)
		foc	Nuga	Watat	pst.rem	sell	c.def
		‘NUGA, Watat betrayed’.

Keupdjio (2020, sec. 2.2) provides convincing evidence that the ex situ variant exhibits core properties of A’ focus movement, such as reconstruction, crossover effects, and island sensitivity. Note that the focus-moved ex situ variant in (5b) is only licit with the pre-focal á marker, a fact to which we will return in Section 2.2. In addition, the focus-fronting construction (5b) exhibits a distinctive HL-tonal pattern on the auxiliary and verb, which Keupdjio (2020) analyses as a morphosyntactic reflex of A-bar movement. More importantly, the ex situ variant in (5b) features the clause-final definite determiner lá, which is also obligatorily found at the end of Mǝ̀dʉ́mbà’s relative clauses (Kouankem 2012). This structural resemblance with relative clauses suggests at least a diachronic relation with focus-partitioning cleft clauses.2

As for what triggers the ex situ realisation of focus, Keupdjio (2020) argues sensu Kiss (1998) and Horváth (2010) that focus fronting is driven by a semantic factor in that it serves to express focus exhaustivity. In this paper, we take issue with Keupdjiou’s analysis of focus left-dislocation in terms of focus exhaustivity. We will show that the exhaustivity inference observable with Mǝ̀dʉ́mbà’s focus is already expressed as part of the lexical meaning of the pre-focal focus marker á, both in situ and ex situ (see Hartmann and Zimmermann (2007a) on Hausa (Chadic) nee/cee, Renans (2016) on Ga (Kwa) ni, Fominyam and Šimík (2017) on Grassfields Bantu Awing lɛ, and Kiemtoré (2022) on Jula (Mande) ale, among others, for a discussion of morphological focus markers that trigger comparable EXH- inferences. For Mǝ̀dʉ́mbà, it follows that the need to express focus exhaustivity is not the driving force behind focus left-dislocation, and that there must therefore be another reason for it. We will show that A’- focus fronting triggers an existence presupposition, as is also the case with English it- clefts (Horn 1981; Rooth 1996; Zimmermann and Onéa 2011; Zimmermann et al. 2020) and with focus fronting in the Kwa languages Akan and Ga (Grubic et al. 2019). Next to this, focus fronting may have additional discourse-pragmatic effects, e.g., additional emphasis on the focus constituent, where emphasis can be understood in terms of discourse highlighting (see, e.g., Gundel’s (1988) First Things First principle or Givón’s (1988) principle Attend to the most urgent task first), mirativity (DeLancey 1997; Peterson 2016), or discourse unexpectedness (Zimmermann 2008; Destruel and Velleman 2014).

The rest of this article is structured as follows: Section 2 provides some background information on Mǝ̀dʉ́mbà and takes a look at the syntax of in situ focus marking with á. We will argue against an analysis in terms of a lower focus projection at the left edge of vP (Tuller 1992; Belletti 2004; i.a.). We will also sketch an analysis of predicate doubling with verb focus, which builds on the assumption that morphological focus marking is often restricted to non-verbal environments (Hartmann and Zimmermann 2007b, 2009; Zimmermann 2016). In Section 3, we analyse the information-structural function of in situ focus marking with á in Mǝ̀dʉ́mbà as marking contrastive focus. We will moreover present a unified analysis of unmarked information focus and marked contrastive focus in this language within a Roothian alternative semantics. This unified analysis also accounts for the optionality in focus marking observed in (4). In Section 4, we turn to the analysis of A’- focus movement in Mǝ̀dʉ́mbà and its conditioning factors. We will show that two common factors that have been argued to trigger focus fronting in other languages—namely, contrast and exhaustiveness inferences—are not the driving forces behind focus fronting in Mǝ̀dʉ́mbà, as these inferences are already part of the lexical meaning of the focus marker á (see also Ylinärä et al. (2023) for evidence from Italian and Finnish that focus fronting is not connected to exhaustivity effects). Instead, we will show that focus fronting in Mǝ̀dʉ́mbà triggers an existence presupposition, the same as found with syntactic focus–background bipartitions in many other languages.

2. The Syntax of In Situ Focus Marking in Mǝ̀dʉ́mbà: Local Attachment to [-V] XPs

Bamileke Mǝ̀dʉ́mbà is a Grassfields Bantu tonal language spoken in Western Cameroon. As shown in (1), the basic word order is S-TAM-VOX, where X stands for any additional prepositional object or adjunct. Mǝ̀dʉ́mbà is a graded tense language (Comrie 1985), with remote and near tense and aspect markers preceding the verb (Nganmou 1991; Kouankem 2012; Mucha 2015, 2017; Mucha and Zimmermann 2016; Keupdjio 2020)3; see Kouankem (2012) for a detailed description, and Mucha (2017) for an in-depth formal analysis of the graded tense system of Mǝ̀dʉ́mbà in interaction with aspect and modality. Keupdjio (2020) provides a detailed analysis of A’- movement, its tonal reflexes on the clausal remnant, and the interactions between A’- fronting and TAM- marking in the language. The Bamileke Mǝ̀dʉ́mbà data reported here come from the second author’s intuitions as a native speaker, as well as from phone elicitations with up to eight other speakers.

2.1. Morphological Focus Marking on Subject, Object, and Adjuncts

With subject focus, the morphological focus marker precedes the focused constituent in its canonical preverbal position. The focus maker can occur with proper nouns (6a), with pronouns (6b), and with full lexical DPs alike, cf. (2a) above.4

(6)	a.	#(á)	ŋgmí	náʔ	sǝ̀ʔǝ̀
		foc	Ngami	pst.rem	come
		‘NGAMI came’. (… and not Numi)
	b.	#(á)	yí	fǝ̀	nɛ́n	ntάnǝ̀
		foc	she	pst.near	go	market
		‘SHE went to the market’. (…and not you)

As already shown in (2b), focus on direct objects is typically expressed by the focus marker á immediately preceding the focused constituent. This holds for full DPs (2b) and pronouns alike, cf. (7). The focus counterpart of (1) in (8) shows the realisation of in situ focus on indirect prepositional objects.

(7)	nyú	fǝ̀	ntʃɔ̀g	á	mǝ̀
	snake	pst.near	bite	foc	me
	‘A snake bit ME (…and not NGAMI)’.
(8)	nnά	nàʔ	fά	tʃàŋ	á	nǔm	nmí
	Nana	pst.rem	give	food	foc	to	Nami
	‘Nana gave food to NAMI (…and not to NGAMI)’.

Finally, adjuncts in Mǝ̀dʉ́mbà can be focus-marked by á as well. Compare the all-new sentence (9a), without narrow focus on any sub-constituent of the clause, with its contrastive focus counterpart in (9b) with the focus marker á; (10a,b) shows the same for instrumental adjuncts:5

(9)	a.	nnά	àʔ		tʃàg	nɛ́			bάn	dàmnʒʉ́
		nana	fut		tom	cook			fufu	tomorrow
		‘Nana will cook fufu tomorrow’.
	b.	nnά	àʔ		tʃàg	nɛ́			bάn	á	dàmnʒʉ́
		nana	fut		tom	cook			fufu	foc	tomorrow
		‘Nana will cook fufu TOMORROW (… and not later TODAY)’.
(10)	a.	bù	àʔ	ʒʉ́	bǔ		dùʔ
		they	fut	eat	with		spoon
		‘They will eat with a spoon’.
	b.	bù	àʔ	ʒʉ́	á		bǔ	dùʔ
		they	fut	eat	foc		with	spoon
		They will eat with a SPOON (… and not with their HANDS)’.

The next section will adduce more evidence to the effect that morphological focus marking is the primary strategy for realising contrastive focus in Mǝ̀dʉ́mbà.

2.2. Arguments for Morphological In Situ Focus Marking in Mǝ̀dʉ́mbà

In addition to morphological focus marking, focused non-subjects can also be realised ex situ in the left-peripheral position of the clause. This was already shown in (5) above. Even though the focused constituent more frequently remains in its canonical position (in situ), both strategies are readily attested in Mǝ̀dʉ́mbà. The ex situ variant of the in situ focus in (8) is shown in (11).6

(11)	á	nǔm	nmí	nnά	náʔ	fά	tʃàŋ	lá
	foc	to	Nami	Nana	pst.rem	give	food	c.def
	‘To NAMI, Nana gave food (…and not to NGAMI)’.

Crucially, the ex situ focus structures in (5) and (11) will be ungrammatical if the morphological focus marker is dropped. From this, we can conclude that the primary strategy of expressing contrastive focus in Mǝ̀dʉ́mbà consists in the insertion of the morphological marker á; see Hartmann and Zimmermann (2009) for parallel facts with respect to Gùrùntùm (West Chadic).7 We will turn to the discourse-semantic motivation for using ex situ focus over the default in situ focus strategy in Section 4.

Another argument for the primary status of in situ focus is that it can be used to express contrastive focus on all kinds of sentential constituents, including adjuncts, whereas focus fronting is blocked with framing adjuncts providing spatial, temporal, manner, or causal information (Keupdjio 2020, p. 66), as (12) from Keupdjio (2020, p. 66, ex. 93b) illustrates.8 The ungrammaticality of (12) may provide support for claims by Frascarelli (2017) that focus and framing are different IS- categories with different properties. We will briefly return to this point in Section 4.3.

(12)	* á	ŋgɣuʔ-muʔ	Nuŋgɛ̀	náʔ	nʤʉ̂n	Watɛ̀t	lá
	foc	year-other	Nuga	pst.rem	see	Watat	c.def
	INTENDED: ‘It was last year that Nuga saw Watat’.

The conclusion that the in situ focus marking with á is the basic focus-marking strategy in Mǝ̀dʉ́mbà receives further support from a closer look at the syntactic distribution of in situ focus constituents. The central finding here is that subject and non-subject term foci always occur in their canonical surface position, irrespective of the presence of the focus marker á.9 For illustration, consider again the all-new base sentence in (1), repeated below, with its counterparts marked for subject focus (13a), direct object focus (13b), and focus on the prepositional indirect object in (8). As can be seen, the presence of focus marking does not result in a change in word order, and the only indication of focus is the particle á preceding the focused constituent. The same holds for the relative positioning of in situ focused temporal, spatial, and other adjuncts, not shown here.

(1)	nnά	náʔ	fά	tʃàŋ	nǔm	nmí
	Nana	pst.rem	give	food	to	Nami
	‘Nana gave food to Nami’.

(13)	a.	á	nnά	náʔ	fά	tʃàŋ	nǔm	nmí
		foc	Nana	pst.rem	give	food	to	Nami
		‘NANA gave food to Nami’.
	b.	nnά	náʔ	fά	á	tʃàŋ	nǔm	nmí
		Nana	pst.rem	give	foc	food	to	Nami
		‘Nana gave FOOD to Nami’.

In the next section, we provide evidence against an alternative analysis of Mǝ̀dʉ́mbà’s in situ focus in terms of movement to a designated high- or low-functional projection, FocP, which would be masked by subsequent movement operations, e.g., movement of V to T.

2.3. No Low-Focus Projection at the Edge of vP

In this section, we argue that a syntactic analysis of focus in terms of designated high- and low-focus projections in the syntax cannot account for the distribution of á-focused constituents in Mǝ̀dʉ́mbà. We focus on Collins and Essizewa’s (2007) concrete proposal for the Kɛwɛ dialect of Kabiye (Gur, Niger–Congo) as a characteristic representative of cartographic analyses (Rizzi 1997). The central idea consists in the assumption of a sentence-internal focus projection at the edge of vP/VP (see, e.g., Tuller (1992), Kayne (1998), Belletti (2004), Kandybowicz (2008), or Aboh (2009) for proposals along these lines). Collins and Essizewa (2007, pp. 193–94) argue that there are two different syntactic positions for focused constituents in Kabiye. The central empirical argument for this claim is that a focused term can be realised either in a sentence-internal position following the main verb, or ex situ in a sentence-initial position. In the former case, the focus is followed by the focus marker na, cf. (14a). In the latter case, the fronted focus exhibits final vowel lengthening of ɛ, cf. (14b), which Collins and Essizewa (2007) analyse as indicating the presence of a covert Foc-head. Crucially, the focus marker na cannot co-occur with the fronted ex situ focus in the sentence-initial position, cf. (14c), nor does the post-verbal focus exhibit lengthening.

(14)	Q:	What language do you understand?
	a.	ma- nɪ́-ʊ		kabiyɛ	na
		1sg-understand- impf		Kabiye	foc
		‘I understand KABIYE’.
	b.	kabiyɛ-ɛ	má-nɪ-́ʊ
		Kabiye-foc	1sg-understand- impf
		‘I understand KABIYE’.
	c.	*kabiyɛ-ɛ	na	má-nɪ-́ʊ
		Kabiye-foc	foc	1sg-understand-impf

Based on the incompatibility of lengthening and na in (14c), Collins and Essizewa (2007) propose that Kabiye has two focus projections, the higher (ex situ) one of which is headed by a null element ∅, while the lower, sentence-internal one is headed by na, cf. (15); (16a,b) show the corresponding structures for sentence-internal and sentence-initial object focus in (14a,b), respectively.

(15)	[FocP  __ [ ∅FOC  [IP V-I [FocP  __ [ naFOC [VP … <V> …]]]]]]
(16)	a.	[IP ma [ [I nɪ́-ʊ] [FocP kabiyɛFOC [ naFOC [VP <ma> < nɪ́> <kabiye>]]]]]
	b.	[FocP kabiyɛFOC [ɛFOC [IP ma [ [I nɪ́-ʊ]  [VP <ma> < nɪ́> <kabiye>]]]]]

The focused object DP in (16a) undergoes short movement to the specifier position of the lower FocP. As the verb subsequently moves to I in order to check off its own feature specifications, the focused object in the lower FocP will end up in a post-verbal surface position. Focused subjects, in contrast, first move to the specifier position of the lower FocP headed by na. From there, they move on to SpecIP, stranding the focus head na in a post-verbal position, cf. (17ab):

(17)	a.	Q:	Who understands Kabiye?
		A:	ɛsɔ ́	nɪ́-ʊ	na	kabiyɛ
			Esso	understand-impf	foc	Kabiye
			‘ESSO understands Kabiye’. (reply to ‘Who understands Kabiye?’)
	b.	[IP ɛsɔ ́ [ [I nɪ́-ʊ] [FocP <ɛsɔ ́>			[ naFOC [VP <ɛsɔ ́> <nɪ́> kabiye]]]]]

Collins and Essizewa (2007) briefly entertain and reject an alternative analysis of focus in Kabiye, in which the focused constituent and the post-focal marker na would form a constituent, as in (18). This alternative does not involve syntactic movement to a designated focus position. Instead, focus would be directly expressed on the in situ focused constituent.

(18)	[ XPFOC -na ]

The authors give two empirical arguments against this analysis: First, in Kabiye, na never occurs with ex situ focused constituents in the sentence-initial position, but this should be possible based on the analysis in (18). Second, the obligatory post-verbal occurrence of na with the focused subject in (17a) suggests that the focused constituent and the post-verbal focus marker do not form a local constituent.

It is easy to see that Mǝ̀dʉ́mbà differs from Kabiye on both counts. First, it only has a single morphological focus marker á, which occurs with instances of in situ and ex situ focus alike. Second, post-verbal stranding of the focus marker with instances of subject focus results in ungrammaticality, cf. (19):

(19)	Q:	Who cooked plantains?
	A:	* nnά	náʔ	nɛ́	á	kǝ̀lò
		Nana	pst.rem	cook	foc	plantains

We can conclude that the focus marker á and focus expressions do indeed form a constituent in Mǝ̀dʉ́mbà. This conclusion is backed up by additional evidence. For one, the linear order of the focused constituent and focus marker is FM >> XP_FOC, and not XP_FOC >> FM. The latter order would be expected, though, if á were the head of FocP, attracting the focused direct-object pronoun to its specifier position to the left; (20b) shows the unattested surface order with FocP.

(20)	a.	nyú	fǝ̀	ntʃɔ̀g		á		mǝ̀
		snake	pst.near	bite		foc		me
		‘A snake bit ME’
	b.	*nyú	fǝ̀	ntʃɔ̀g	[FocP mǝ̀		[áFoc […

Finally, recall from (8) and (13) above that Mǝ̀dʉ́mbà exhibits no evidence whatsoever for any focus-induced changes in word order in the post-verbal domain. Indeed, reordering the á-marked focused indirect object in (8) to a position preceding the direct object incurs a grammaticality violation, as shown in (21):

(21)	a.	*nnά	náʔ	fά	[á	nǔm	nmí ]	tʃàŋ
		Nana	pst.rem	give	foc	to	Nami	food

In light of the available evidence, we therefore conclude that focus on nominal terms such as DP/PP- arguments and adjuncts in Mǝ̀dʉ́mbà is not licensed by syntactic movement to some lower FocP. Hence, there is no evidence for the projection of low-focus phrases in the syntax of Mǝ̀dʉ́mbà and, consequently, the focus marker á does not originate in a syntactic head (Foc) position. Instead, as in other morphologically focus-marking languages (Hartmann and Zimmermann 2009; Zimmermann and Onéa 2011; Assmann et al. 2023), the morphological focus marker á can be considered a direct functional counterpart to focus accenting in intonation languages. It does not have a syntactic function, but only the information-structural function of making the focus constituent prominent. As such, it is base-generated as a syntactic constituent with the term focus in situ through external merging (Chomsky 2001), corresponding to the structure in (18) above: [_DP/PP á [DP/PP]]. This analysis accounts directly (i.) for the syntactic distribution of focused constituents, (ii.) for the obligatory presence of the á marker in all contrastive focus contexts, and (iii.) for the linear order of the focus marker and focused constituent. The occurrence of moved ex situ foci in the sentence-initial position must be motivated by other factors, then, the same as with focus–background clefting in English. The next section extends the analysis of focus in Mǝ̀dʉ́mbà to instances of verb focus, which also requires the presence of the focus marker á, but now on a copy of the verb.

2.4. The Formal Realisation of V- Focus: Verb Doubling Plus á Marking

In the remainder of this section, we add some information on the realisation of V- focus in Mǝ̀dʉ́mbà, leaving a formal syntactic analysis for another occasion. From a discourse-semantic perspective, focus on verbal predicates evokes alternatives to the action, event, or state denoted by the verb or VP. By presenting the verbal meaning against the background of focus alternatives, verb focus puts emphasis on the action undertaken, or the situation obtained, against the background of alternative actions or situations that are conceivable in the context. The interpretive effect of V(P)- focus is informally given in (22) in the form of a pseudocleft:

(22)	What SUBJ did (with X) was V

The formal realisation of verb focus in Mǝ̀dʉ́mbà proceeds in a similar fashion to term focus marking, but with one additional complication: as shown in (3a) above, verb focus is expressed by placing the by now-familiar focus marker á on a second occurrence of the verb (=V_COPY) in the post-verbal domain. With intransitive verbs, á + V_copy directly follows the doubled verb, cf. (23). With transitive verbs, á + V_copy must follow the direct object NP, cf. (3a) repeated as (24a). Placing á + V_copy directly after the verb results in ungrammaticality, cf. (24b). With both verb types, tense and aspect markers precede only the first instance of the verb, which we treat as the full finite form of the verb bearing (abstract) specifications for finiteness and agreement.

(23)	a.	á	kʉ̀b
		he	cut
		‘He has cut’.
	b.	á	kʉ̀b	á	kʉ̀bǝ̀
		he	cut	foc	cut
		‘He has CUT (he has not eaten)’.
(24)	a.	nnά	náʔ	lùʔ	bí	á	lùʔù	[V DO á-VCOPY]
		Nana	pst.rem	take	knife	foc	take
		‘Nana TOOK the knife (…she did not steal it)’.
	b.	*nnά	náʔ	lùʔ	á	lùʔù	bí	[V á-VCOPY DO]

Below, we will argue that the complex á + V_copy is a constituent located at the right edge of the VP/vP projection, the same as with non-verbal adjuncts. To this end, we will discuss two central properties of verb doubling: First, an identity constraint on the copy suggests that the copying mechanism takes place in the syntactic component. Secondly, the syntactic distribution of á + V_copy resembles that of non-verbal adjuncts, suggesting that the categorical status of V_copy may also be [-V], or at least [-finite]. Assuming head movement of the finite verbal head from V to v to T, the resulting structure would resemble parallel verb chain formation in Gungbe, as discussed by Aboh and Dyokanova (2009), with the sole difference being that Mǝ̀dʉ́mbà does not have a designated focus projection in the left-clausal periphery. In contrast, the available data on verb doubling in Mǝ̀dʉ́mbà are in line with an analysis in which the focus-marked verb copies have non-verbal or non-finite traits and are freely adjoined in the post-verbal domain, the same as all other non-verbal adjuncts.

First, V-copy formation in Mǝ̀dʉ́mbà is subject to a constraint of strict lexical identity: the verbal copy marked by the focus marker á must be an exact lexical copy of the finite verb. This is illustrated in (25) and (26). The examples in (25) are licit, as each involves copies of the same verb root, làb and vʉ́l ‘to beat’. By contrast, the examples in (26) show that the finite verb and the second verbal element carrying the focus marker cannot involve different lexical roots, not even if even if the two verbs are (near) synonymous:

(25)	a.	nnά	náʔ	làb	í	á	làbǝ̀
		Nana	pst.rem	beat	3sg	foc	beat
		‘Nana BEAT him’.
	b.	nnά	náʔ	vʉ́l	í	á	vʉ́lǝ̀
		Nana	pst.rem	beat	3sg	foc	beat
		‘Nana BEAT him’.

(26)	a.	*nnά	náʔ	làb	í	á	vʉ́lǝ̀
		Nana	pst.rem	beat	3sg	foc	beat
		INTENDED: ‘Nana BEAT him’.
	b.	nnά	náʔ	vʉ́l	í	á	làbǝ̀
		Nana	pst.rem	beat	3sg	foc	beat
		INTENDED: ‘Nana BEAT him’.

The lexical identity constraint points to a syntactic relation between the main verb and its copy: the two verbal parts do not seem to originate as separate lexical items in the numeration (Chomsky 1995). Rather, the observed copying seems to be the result of syntactic movement plus spellout of the base.

Secondly, at least in some instances, the verbal copy under verb focus must be accompanied by the infinitival marker nʉ̀. This is the case, for instance, with instances of exclusive verb focus with the exclusive particle ndà^ʔ ‘only’ in (27), which appears to select for nominal complements only; cf. Hartmann and Zimmermann (2007b, 2009) for comparable facts in Chadic Tangale and Gùrùntùm:10

(27)	nnά	náʔ	lùʔ	bí	á	ndàʔ	nʉ̀	lùʔù
	Nana	pst.rem	take	knife	foc	only	inf	take
	‘Nana only TOOK a knife (she didn’t use it)’.

In their infinitival guise, Mǝ̀dʉ́mbà verbs show the categorial behaviour of nouns or non-verbal elements in that they are licit in typical NP- positions, e.g., with demonstratives (28ab) and with possessive markers (29ab) (see also Frascarelli and Ramaglia (2014) for an analysis of such structures in Italian):

(28)	a.	yə̂n	mɛ́n	lì
		this	child	here
		‘this child’
	b.	yə̂n	nʉ̀	làb	lì
		this	inf	beat	here
		‘this (way of) beating’
(29)	a.	yù	mɛ́n	yə̂n	lá
		your	child	that	there
		‘that child of yours’
	b.	yù	nʉ̀	ʒʉ́	tʃàŋ	yə̂n	lá
		your	inf	eat	food	that	there
		‘that your (way of) eating’.

The [-V] categorial status of the á + nʉ̀-marked copy of the lexical root in (27) thus resembles that of other adjunct material in the post-verbal domain. We therefore assign it the category feature [-V], which is found with NPs and PPs. We could extend the [-V] analysis to instances of á + V_COPY without nʉ̀, as in (25a,b), but in the absence of positive evidence on the categorial status of the bare verbal copies we will delegate this question to another occasion.

More importantly for our purposes, distributional facts suggest that the á-marked copy of the verb root is right-adjoined to VP/vP in the post-verbal domain, the same as all other post-verbal adjuncts. First, it can be shown that the rightward position of á + V_COPY is structurally located below NEG and Q: á + V_COPY must take semantic scope below negation (30a), and á + V_COPY must precede the clause-final question marker kí (30b). Moreover, verb doubling is also licit in relative clauses and embedded complement clauses, cf. (31a,b). This shows that the position of á + V_COPY is not restricted to matrix clauses. Rather, á + V_COPY is realised in a position right-adjoined to its local VP/vP domain.

(30)

a.

nnά

náʔ

kʉ̀

lùʔ

bí

á

lùʔù

Nana

pst.rem

neg

take

knife

foc

take

‘Nana did not TAKE a knife’.

not: ‘What Nana didn’t do with the knife was to take it’.

b.

nnά

náʔ

làb

í

á

làb

kí?

Nana

pst.rem

beat

3sg

foc

beat

q

‘Did Nana BEAT him?’

(31)

a.

mɛ́n

yə̂n

lá

[ zǝ̀

nnα

náʔ

làb

í

á

làb

lá]

child

this

here

rel

Nana

pst.rem

beat

3sg

foc

beat

def

‘the child that Nana BEAT’

b.

mǝ̀

fǝ̀

kwɛ̌dǝ̀

ŋkɔ̀g

[bʉ̀

à

kò

í

á

kǒ]

1sg

pst.near

think

yesterday

that

3sg

like

3sg.obj

foc

like

ndô

mǝ̀

yʉ̀n

[bʉ̀

nmí

bɔ̌

nnά

á

bɔ̌]

now

1sg

see

that

Nami

hate

Nana

foc

hate

‘I thought yesterday that she LIKES him, NOW I think that Nami HATES Nana’.

The following data points show that á + V_COPY must occur in the post-verbal domain following the finite verb (32a–c), and that the á marker must be on the non-verbal copy (32d).

(32)	a.	*nnά	náʔ	á	lùʔù	lùʔ	bí	[*á-VCOPY > VFIN]
		Nana	pst.rem	foc	take	take	knife
	b.	*nnά	á	lùʔù	náʔ	lùʔ	bí	[*á-VCOPY > T]
		Nana	foc	take	pst.rem	take	knife
	c.	*nnά	náʔ	á	lùʔù	bí	lùʔ	[*á-VCOPY DO > VFIN]
		Nana	pst.rem	foc	take	knife	take
	d.	*nnά	náʔ	á	lùʔ	bí	(lùʔù)	[* á-marking on VFIN]
		Nana	pst.rem	foc	take	knife	take

Moreover, V doubling does not result from leftward V movement to a higher v projection plus focus marking on the lower verbal copy spelled out in the base position, cf. (33). Based on the analysis in (33), it would be the first instance of the verb that is copied from the second.

(33)	Impossible structure for V-doubling in Mǝ̀dʉ́mbὰ:
	SUBJ [vP <VCOPY>+v … [VP OBJ á-<V>]]

Evidence against this analysis comes from the possibility of doubling lower verbs in Mǝ̀dʉ́mbà serial verb constructions (SVCs). According to Aboh (2009), such SVC structures involve the base generation of V1 in a light verb projection v. This is shown in (34), which is structurally comparable to (33).

(34)	SVC-structure in Aboh (2009):
	SUBJ [vP V1+v … [VP OBJ V2]]

But this analysis would leave the possible doubling of V3 in (35a) and V2 (and V1) in (35b) unaccounted for. With the high light verb projection filled by V1, there simply is no space to accommodate any other focus copies of the verb.

(35)

a.

bìn

àʔ

nɛ́n

ntάnǝ̀

n-dùʔ

mbàb

n-sǝ̀ʔ

yí

á

sǝ̀ʔǝ̀

2pl

fut

go

market

N-take

meat

N-come

3sg.inan

foc

come

‘You will go to the market and BRING the meat’.

b.

nnά

náʔ

lùʔ

ŋwàʔnì

á

lùʔù

n-nɛ́n

yí

mά

ndà

á

nɛ́nǝ̀

Nana

pst.rem

take

book

foc

take

N-go

3sg.inan

to

house

foc

go

‘Nana TOOK the book home’.

The SVC data thereby show that the first verb occurrence in verb doubling under focus does not occupy a high light verb projection, and that the second occurrence of the verb is not in its syntactic base position. Rather, it is the focus-marked second verb occurrence that appears to be a copy in a derived position in each subpart of the SVC.

The final relevant dataset shows that á + V_COPY has the same syntactic distribution as other post-verbal adjuncts. In particular, á + V_COPY is interchangeable with other post-verbal adjuncts, which themselves exhibit variable word order:

(36)

a.

nnά

lú

fά

tʃàŋ

nǔm

nάmí

ŋgàb

mùʔ

dʒʉ̀ʔ

ŋwàʔnì

á

fά

Nana

pst.near

give

food

to

Nami

week

last

at

school

foc

give

‘Nana GAVE food to Nami last week at school’.

TEMP > LOC > VCOPY

b.

nnά

lú

fά

tʃàŋ

nǔm

nάmí

dʒʉ̀ʔ

ŋwàʔnì

ŋgàb

mùʔ

á

fά

Nana

pst.near

give

food

to

Nami

at

school

week

last

foc

give

‘Nana GAVE food to Nami last week at school’.

LOC > TEMP > VCOPY

c.

nnά

lú

fά

tʃàŋ

nǔm

nάmí

á

fά

dʒʉ̀ʔ

ŋwàʔnì

ŋgàb

mùʔ

Nana

pst.near

give

food

to

Nami

foc

give

at

school

week

last

‘Nana GAVE food to Nami at school last week’.

VCOPY> TEMP > LOC

d.

nnά

lú

fά

tʃàŋ

nǔm

nάmí

dʒʉ̀ʔ

ŋwàʔnì

á

fά

ŋgàb

mùʔ

Nana

pst.near

give

food

to

Nami

at

school

foc

give

week

last

‘Nana GAVE food to Nami last week at school’.

TEMP > VCOPY> LOC

As already mentioned above, these findings are compatible with an analysis in which á + V_COPY is a non-verbal category, which right-adjoins after movement to its local VP/vP, the same as other (spatial, temporal, instrumental, etc.) adjuncts:

(37)	[vP Subj [VP [VP V DO] … [á + VCOPY] ] …]]

With these observations in place, we will leave a detailed formal analysis of verb doubling under focus in Mǝ̀dʉ́mbà for another occasion. In what follows, we will concentrate on focus marking on clearly non-verbal DP and PP terms.

3. Information Structure: Á Marks Contrastive Focus

This section addresses the information-structural import of morphological focus marking with á. First, we will show in more detail that Mǝ̀dʉ́mbà marks new information focus and contrastive focus (Kiss 1998; Kratzer and Selkirk 2020) in different ways. The minimal pair in (38a,b) shows that Mǝ̀dʉ́mbà has no special morphological or syntactic marking for open information focus (38a), whereas the presence of á marks contrastive focus, as indicated by the presence of explicit alternatives in the context (38b):

(38)	a.	Q:	What could Nana possibly have taken?						New information focus: no á
			nnά	náʔ	lùʔ	bí
			Nana	pst.rem	take	knife
			‘Nana took a knife’.
	b.	Q:	Did Nana take [a knife or a pen]?						Contrastive focus: á
			nnά	náʔ	lùʔ	#(á)	bí
			Nana	pst.rem	take	foc	knife
			‘Nana took a KNIFE’.

The central claim of this section is that the basic information-structure (IS) function of explicit focus marking with á in Mǝ̀dʉ́mbà is to express the presence of contextually salient contrasting alternatives in the discourse. This IS category has been alternatively labelled as focus (Rooth 1992; Krifka 2008), FoCus (Kratzer and Selkirk 2020; Katz and Selkirk 2011) or contrastive focus (Neeleman et al. 2009). Drawing on elicitations and the second author’s intuitions as a native speaker, we also show that the focus marker á must occur in typical contrastive focus environments. We will then point to some instances of apparent optionality in the placement of á. Finally, we will sketch a unified formal analysis of information and contrastive focus in the spirit of Rooth (1992), which accounts for the basic pattern as well as for the observed optionality in the placement of á.

3.1. The Facts: Á Is Required in Contrastive Focus Environments

The Mǝ̀dʉ́mbà focus marker á occurs in all discourse environments that trigger contrastive focus, such as (i.) in answers to alternative questions, (ii.) in answers to D-linked wh- questions with a restricted question domain, in (iii.) corrective and (iv.) counter-presuppositional assertions, and (v.) with the exclusive particle ndà^ʔ ‘only’.

The inherent contrastive nature of á has already become apparent in many of the paraphrases for the á examples in Section 2, in which the contrasting alternatives were explicitly mentioned in the subsequent context. This is shown again in (8), repeated.

(8)	nnά	nàʔ	fά	tʃàŋ	á	nǔm	nmí
	Nana	pst.rem	give	food	foc	to	Nami
	‘Nana gave food to NAMI (…and not to NGAMI)’.

Secondly, á is obligatory in answers to disjunctive questions, as shown immediately above in (38b). Disjunctive questions are one of the most reliable contexts for triggering contrastive focus (see also: discussion and examples in Rooth (1996)).

Thirdly, á is obligatory in answers to D-linked wh- questions, which come with a restricted question domain in the form of a set of salient alternatives (Katz and Selkirk 2011). We have already encountered an example in (4b) above, in which á was obligatory in both the wh- question and answer for reasons of discourse felicity. This is expected if the basic function of á consists in pointing to the presence of contextually salient alternatives; (39) illustrates the same point:

(39)

Q:

#(á)

wʉ́

nʉ̀tɛ́t

nnά,

nǔmí

bǔ

nmí

bɛ̀n

náʔ

nʒʉ́

tò

á

foc

who

between

Nana,

Numi

and

Nami

rel

pst.rem

win

game

q

‘Who among Nana, Numi and Nami won the game?’

A:

#(á)

nnά

bɛ̀n

náʔ

nʒʉ́

tò

foc

nana

rel

pst.rem

win

game

‘NANA won the game’.

Recall from above that á is not required in (answers to) wh- questions that ask for plain new information in the absence of contextually salient alternatives, a point to which we return.

Next, á occurs in corrective assertions, cf. (40B), where the presence of á points to a speaker–addressee disagreement concerning the focused part of the utterance (Zimmermann 2008):

(40)	A:	nǔmí	nɛ́	kǝ̀lò
		Numi	cook	plantains
		‘Numi has cooked plantains’.
	B:	ŋgɛ́	nǔmí	nɛ́	#(á)	bǝ̀lɔ̀ŋ
		no	Numi	cook	foc	potatoes
		‘No, Numi has cooked POTATOES’.

The contrastive focus nature of á is further demonstrated by its occurrence with instances of counter-presuppositional focus. These occur whenever the speaker sees the need to address and repair an incorrect background assumption of the addressee (Destruel and Velleman 2014). This happens, e.g., if the discourse interlocutor puts forward some factually incorrect information as uncontested or given. For illustration, consider the wh- question in (41Q), in which the questioner seems to take it for granted that Numi has cooked plantains; here, the discourse repair in (41A) must contain the focus marker á.

(41)	Context: Numi cooked potatoes.
	Q:	When did Numi cook plantains? ⇒ BACKGROUNDED: Numi cooked plantains.
	A:	nǔmí	náʔ	nɛ́	#(á)	bǝ̀lɔ̀ŋ
		Numi	pst.rem	cook	foc	potatoes
		‘Numi cooked POTATOES!’

Finally, á is obligatory in the presence of the exclusive focus particle ndà^ʔ ‘only’, cf. (42). Exclusive particles are standardly analysed as inherently associating with contrastive focus, since their core semantic function consists in the exclusion of (salient) alternatives to the focus denotation (Beaver and Clark 2008; Katz and Selkirk 2011).

(42)	nǔmí	kʉ́	swɛ̀n	*(á)	ndàʔ	bànànà	nvʉ́	ntάnǝ̀
	Numi	prog	sell	foc	only	banana	at	market
	‘Numi sells only BANANAS at the market’.

The obligatory co-occurrence of ndà^ʔ and á follows directly from the nature of á as a contrastive focus marker that points to the presence of contextually salient alternatives. In sum, then, this section provides conclusive evidence for the analysis of the Mǝ̀dʉ́mbà focus marker á as a marker of contrastive focus; it presents and singles out the focused information against the background of a set of contextually salient alternatives.

3.2. Interpretation of New Information vs. Contrastive Focus: Towards a Unified Analysis

As with many other West African languages from other families (e.g., Hartmann and Zimmermann 2007c; Grubic et al. 2019; i.a.), Mǝ̀dʉ́mbà allows for a certain amount of optionality when it comes to the placement of the focus marker á in wh- questions and their corresponding answers. The relevant data were already presented in (4), repeated here:

(4)

a.

Q:

nnά

náʔ

fά

tʃàŋ

nǔm

wʉ́

(new information: no á-marking!)

Nana

pst.rem

give

food

to

whom

‘To whom did Nana give food?’

A:

nnά

náʔ

fά

tʃàŋ

nǔm

nmí

Nana

pst.rem

give

food

to

Nami

‘Nana gave food to NAMI’.

b.

Q:

nnά

náʔ

fά

tʃàŋ

á

nǔm

wʉ́

(contrastive focus: á-marking!)

Nana

pst.rem

give

food

FOC

to

who

‘To whom/which of them did Nana give food?’

A:

nnά

náʔ

fά

tʃàŋ

#(á)

nǔm

nmí

Nana

pst.rem

give

food

FOC

to

Nami

‘Nana gave food to NAMI (and not to NUMI)’.

As shown in Section 3.1, the presence or absence of the focus marker á in wh- questions depends on their status as D-linked or not, i.e., whether such questions ask about a contextually salient set of alternatives (contrastive focus), or whether they are open, new questions without predefined answer space (new information). If á is present in the question, it must also be present in the answer. If á is absent in the question, it will normally be absent in the answer as well, but its optional occurrence is also licit. The distinction between two subtypes of focus is well known since Kiss’s (1998) seminal work on Hungarian, and it is also attested in many other West African languages, such as Ga (Renans 2016), Ngamo (Grubic 2015), and Akan (Grubic et al. 2019). Theoretically, this raises the question of how to analyse this in formal semantic terms. The first option, advocated by Krifka (2008) and Kratzer and Selkirk (2020), among others, would consist in only analysing instances of contrastive focus as bona fide instances of focus in the sense of Rooth (1992, 1996). Rooth (1992) proposed a two-layer semantic model with an ordinary semantic value 〚 〛 ^o and a focus semantic value 〚 〛 ^f, where the focus value contains the (projected) alternatives to the ordinary value of the focus constituent, including the latter; 〚 〛 ^o- and 〚 〛 ^f- values interact mediated by a covert focus ~ operator, which introduces a contextually bound covert variable C_i. The value of C_i is assignment-dependent and stores all of the contextually salient alternatives at the relevant point in the discourse. The meaning contribution of ‘~’ consists in the addition of the three presuppositions in (43) (with TP the clausal constituent in the scope of ‘~’):

(43)	i.	〚 Ci 〛 g ⊆ 〚 TP〛 f	(Q-A congruence)
	ii.	〚 TP〛 o ∈ 〚 Ci〛 g	(relevance)
	iii.	∃p: p ≠ 〚 TP〛 o ∧ p ∈ 〚 Ci〛 g	(nontrivial salient alternatives)

In a nutshell, (43i) ensures question–answer or, more generally, focus congruence, by requiring that the contextually salient alternatives in C_i be a subset of the alternatives grammatically projected from the focus-marked constituent; (43ii) ensures relevance by requiring that the ordinary meaning of the clause be among the contextually salient live alternatives; (43iii) ensures non-triviality by requiring that C_i contain more than the ordinary meaning to choose from. According to the focus=contrast approaches referred to above, one could postulate that only the Q/A sequences in (4b) with the focus marker á contain the ‘~’ operator and C variable in their structural representation, which, in turn, trigger a Roothian interpretation of focus. By contrast, plain new-information questions and answers without á in (4a) would need to be interpreted some other way, presumably with anupdate semantics on the Common Ground (Stalnaker 1978). The focus marker á would thereby simply function as an overt indicator of the underlying focus semantics.

This analysis is at odds with another prominent approach to the meaning of focus as making direct reference to the question under discussion (QUD) (Roberts 2012) or current question (CQ) (Beaver and Clark 2008). In this more encompassing approach to focus, both Q/A sequences in (4) would involve focus, irrespective of the presence or absence of á. What, then, would be the discourse-semantic function of á in such a system? In a unified Roothian analysis, one could postulate that all QUD-related clauses contain a focus, for which reason they include the ‘~’ operator and context variable C in their structural representation. What differs, and what would be indicated by the presence of á in Mǝ̀dʉ́mbà, is the way in which the C variable is resolved. It seems plausible that this C variable is subject to the same two resolution strategies that are at work with other variables in natural language semantics, namely, contextual binding on the one hand, and existential closure on the other. These two mechanisms have been proposed, e.g., for the choice-function variable CH_f (Reinhart 1997; Kratzer 1998), for the tense variable t (Partee 1973; Ogihara 2011), and for the situation variable s (Zimmermann and Duah 2022). Turning back to the Mǝ̀dʉ́mbà focus cases, we therefore propose that the only semantic difference between (4aA) and (4bA) is whether the C variable is existentially closed, such as in (4aA) without á, cf. (44a), or whether it is contextually bound, such as in (4bA) with á, cf. (44b).

(44)	a.	∃C[TP2 [TP1 nnά náʔ fά tʃàŋ nǔm nmíF]~C]
	b.	[TP2 [TP1 nnά náʔ fά tʃàŋ á nǔm nmíF]~Ci]

Note that the computation of the ‘~’ presuppositions per (43) still depends on a set of structurally derived focus alternatives, even in the absence of explicit focus marking with á. We assume that, in the absence of explicit focus marking, the focus alternatives are derived from a default abstract F marking at the sentential TP level, giving rise to a set of alternative propositions. This is in line with the observation that any new assertion relies on alternatives in the sense that the new information conveyed by the sentence is presented against the background of alternative assertions that could have been made (Matić and Wedgwood 2013). Formally, such abstract F marking can be modelled in the unalternative framework developed by Assmann et al. (2023) for other languages with morphological focus marking, e.g., Gùrùntùm, Buli, and Hausa. The authors propose an elsewhere system according to which there are specific (morphological) focus markers that indicate constituent focus on a smaller focus domain (V, OBJ, VP, SUBJ). If such markers are absent, F marking is automatically placed on the next higher domain, typically VP or TP, for which no explicit formal marking exists in the languages; see Assmann et al. (2023) for details. Applied to Mǝ̀dʉ́mba, this means that F marking will be placed on VP or TP in the absence of á, which only marks narrow term focus or focus on V. Since 〚 TP_F 〛 ^f contains any proposition whatsoever, (43i) is trivially satisfied.

The unified analysis of information focus and contrastive focus in (46a,b) has two major advantages: First, it can directly account for the fact that, in languages with focus/new information accenting, such as English or Hungarian, contrastive focus marking often involves additional marking on top of new information marking, cf. Zimmermann and Onéa (2011). In Hungarian, this extra marking takes the form of syntactic displacement. In English, it comes in the form of a steeper pitch contour (Katz and Selkirk 2011). Secondly, the unified analysis provides a principled account for some optionality in the distribution of contrastive focus marking. Such optionality is found not only in Mǝ̀dʉ́mbà, but also in other African languages exhibiting optional focus marking, e.g., Chadic Hausa (Hartmann and Zimmermann 2007c) and the Kwa languages Ga and Akan (Grubic et al. 2019). In principle, there are two pathways for the optional omission or placement of contrastive focus markers. First, contrastive focus marking may be optionally dropped in contexts with salient alternatives, such as corrections or answer alternatives. This follows, since contrastive focus marking indicates the contextual recoverability of a salient set of alternatives, i.e., the existence of a contextual assignment function g providing a value to C_i, cf. (44b). The existence of such an assignment, however, directly entails the existence of a set of alternatives C, but not vice versa.

(45)	∃g [∃P 〚 Ci 〛 g = P] ⇒ ∃C (namely P)

Conversely, it is also possible to place and interpret a contrastive focus marker in wh- questions and their corresponding answers, even in the absence of salient alternatives in the immediate context. Competent speakers will most often be able to accommodate the existence of a closed set of alternatives C_i in the absence of contextual information, even if the exact value of C_i remains unknown. This accommodation process from an unconstrained question and focus restriction to a restricted one is informally sketched in (46):

(46)	∃C = {x: x∈WH} ⇒ACCOMODATE ∃P⊂{x: x ∈ WH} ∧ ∀x∈P[ x is context. salient]: Ci = P

The accommodation of C_i thereby accounts for the optional occurrence of contrastive focus marking in contexts without salient alternatives. In sum, a unified Roothian analysis of contrastive and new information focus seems feasible, where the difference between the two lies in how the context variable C is bound. This unified analysis has the further advantage that it can account for some optionality in the placement of contrastive focus markers.

3.3. Interim Conclusion

What we have seen so far is that Mǝ̀dʉ́mbà has explicit focus marking in situ in the form of a morphological focus marker á. This focus marker is placed directly in front of the focus constituent, with which it forms a base-generated constituent. We have also seen that this form of morphological focus marking does not involve syntactic movement of A’ to a designated focus position. The information-structural function of the focus marker á consists in the marking of contrastive focus. Contrastive focus points to the existence of a contextually salient set of alternatives, which can be formally modelled in the form of a contextually bound context variable C_i (Rooth 1992, 1996). In sum, then, we have seen that Mǝ̀dʉ́mba does not have to resort to syntactic movement in order to express the basic IS- function of contrastive focus. This brings us to another question: what are the licensing factors behind the optional ex situ occurrence of contrastive foci in the form of focus left-dislocation? We will address this question in the next and final section of the article.

4. What Triggers Focus Left-Dislocation in Mǝ̀dʉ́mbà?

This section discusses the possible reasons behind focus left-dislocation in Mǝ̀dʉ́mbà; (47) illustrates again the ex situ focusing of the direct object, together with its in situ counterpart:

(47)	a.	á	kǝ̀ló	nǔmí	nɛ́	lá	ex situ
		foc	plantain	numi	cook	c.def
	b.	nǔmí	nɛ́	á	kǝ̀ló		in situ
		numi	cook	foc	plantain

We have already seen that, in Mǝ̀dʉ́mbà, the basic function of contrastive focus marking is lexically coded in the morphological focus marker á, so the expression of contrast per se cannot be the driving force behind ex situ focus, pace Cruschina (2012). The literature offers other possible discourse-semantic reasons for focus left-dislocation, such as (i.) the expression of exhaustivity inferences (Szabolsci 1981; Kiss 1998; Horváth 2010; Grubic et al. 2019), (ii.) the triggering of an existence presupposition on the backgrounded part of the utterance, as typically found with cleft-like focus–background partitions (Horn 1981), or the expression of additional emphasis in terms of noteworthiness (Gundel 1988; Givón 1988), mirativity, or discourse unexpectedness (Zimmermann 2008; Destruel and Velleman 2014).

According to Keupdjio (2020), focus left-dislocation serves the interpretive function of expressing an exhaustivity (EXH-) inference, which—according to him—is not attested with the morphological focus marker á in the absence of focus left-dislocation. In this section, we will take issue with this core claim of Keupdjio (2020) on the EXH-driven nature of focus left-dislocation. In Section 4.1, we will provide conclusive evidence to the effect that it is the morphological focus marker á that triggers the EXH- inference with in situ focus and ex situ focus alike. We will further show that Keupdjio’s (2020) arguments against the exhaustivity of á are inconclusive, or even factually incorrect. Mǝ̀dʉ́mbà thereby resembles another Grassfields Bantu language, Awing, for which Fominyam and Šimík (2017) have shown that the morphological focus marker lɛ likewise expresses contrastive focus and exhaustivity as part of its lexical meaning. Moreover, as already noted in the Introduction, not-at-issue EXH- inferences with morphological focus markers seem to be a pervasive feature of West African languages. Parallel effects have been observed, for instance, for Hausa (Chadic, Hartmann and Zimmermann 2007a), Ga (Kwa, Renans 2016), and Jula (Mande, Kiemtoré 2022). Our empirical claims regarding the inherently exhaustive nature of á are based on the intuitive judgments of the second author, which were checked in phone interviews with up to eight other native speakers of Bamileke Mǝ̀dʉ́mbà; see Supplementary Materials.11 In Section 4.2, we will then show that A’ focus fronting in Mǝ̀dʉ́mbà triggers an existence presupposition. Mǝ̀dʉ́mbà thereby resembles other West African languages, such as Akan and Ga (both Kwa, Niger-Congo), for which this effect is likewise attested (Grubic et al. 2019). In Section 4.3, we briefly discuss the possibility that focus fronting serves additional discourse-pragmatic ends, such as the expression of noteworthiness, and we briefly address the question of why focus fronting is not licit with temporal, spatial, or instrumental adjuncts.

4.1. Focus Movement Does Not Express Exhaustivity—But the Focus Marker á Does!

In this section, we provide conclusive evidence that the expression of EXH- inferences is not dependent on the ex situ realisation of focus. Instead, it is already coded in the meaning of the morphological focus marker á. Section 4.1.1 provides the evidence based on a range of well-established diagnostics for EXH- inferences. Section 4.1.2 adds a few comments on the putative evidence against the exhaustive nature of á put forward by Keupdjio (2020). Section 4.1.3 shows that the EXH- inference is not at issue. The findings in this section resonate with the findings of Frascarelli et al. (this paper), which also show that focus fronting is not (necessarily) connected with exhaustivity.

4.1.1. In Situ á Triggers Not-at-Issue EXH- Inferences

We will demonstrate the inherently exhaustive nature of á on the basis of a series of by-now standard exhaustivity tests; see, in particular, Hartmann and Zimmermann (2007a), Grubic et al. (2019), and Beaver and Onea (2009), i.a.: (i.) the infelicity in non-exhaustive contexts, (ii.) the infelicity of additive follow-ups, (iii.) the availability of EXH-driven entailments, and (iv.) and/but- continuations (Beaver and Onea 2009). Throughout, we demonstrate for non-initial object foci in order to rule out the possibility of a hidden cleft structure with sentence-initial focused subjects; cf. FN4.

i. Infelicity in non-exhaustive contexts: In this test, the á-marked focus constituent occurs in a context specifying that the focus information is non-exhaustive. In (48), the context specifies that Ndzumi has cooked two things: plantains and potatoes. The á-marked focus is hence non-exhaustive in only mentioning plantains, and it is infelicitous in this context.

(48)	Context:		Ndzumi, Ngami’s sister has cooked plantains and potatoes. Nana is Ngami’s friend. Nana is hungry and wants to go and eat Ngami’s food with him. He knows that Ndzumi usually does the cooking. He asks:
	N:	dzúmí	nɛ́	á	kʉ́?	Ng:	#dzúmí	nɛ́	á	kǝ̀ló
		Ndzumi	cook	foc	what		Ndzumi	cook	foc	plantains
		‘What has Ndzumi cooked?’					‘Ndzumi has cooked PLANTAINS’.

The nine speakers consulted on (48) commented on the infelicitous status of Ngami’s answer with á as ‘incomplete’, ‘giving half of the information’, ‘withholding information’, or ‘lacking information’, and thereby being infelicitous. We take this as strong evidence for the exhaustive nature of in situ foci with á, consistent with Keupdjio (2020).

Example (49) shows a variant of the test used by Hartmann and Zimmermann (2007a). The only difference from (48) is that the other actual alternatives are not explicitly listed. Again, the answer in (49A) with á is infelicitous (checked with six speakers) because ‘it restricts the things that Numi sells only to bananas’ [speaker comment].

(49)	Context:		Numi is a street vendor and sells a great variety of things. Nana asks: ‘What does Numi sell?’
	A:	# nǔmí	tʃwɛ̌d	swɛ̀n	á	bànànà
		Numi	pres	sell	foc	banana
		‘Numi sells BANANAS’.

Finally, note that the focus-accented counterparts of both examples in English are fine, since accent focus is semantically non-exhaustive. The contrast with Mǝ̀dʉ́mbà shows that in situ foci with á in Mǝ̀dʉ́mbà are exhaustive.

ii. Infelicity of additive follow-ups (Hartmann and Zimmermann 2007a): This widely used test was originally employed to show EXH- inferences with focus fronting in Hungarian (e.g., Kiss 1998). The test checks for the felicity of additive follow-ups in which the original answer (here á kǝ̀ló ‘foc plantain’) is subsequently extended by additional information with the additive particle (here: á mbà bǝ̀loŋ ‘foc also potatoes’). Again, such additive follow-ups are infelicitous in the presence of the focus marker á, contrasting with the felicitous English paraphrase with non-exhaustive accent focus. The nine speakers commented on the infelicitous follow-up as ‘incomplete’, ‘unclear’, ‘incomprehensible’, or ‘non-exhaustive’. Again, this points to an inherent semantic exhaustivity of in situ focus with á.

(50)	Context:		Numi, Ngami’s sister has cooked plantains and potatoes. Nana is Ngami’s friend. Nana is hungry and wants to go and eat Ngami’s food with him. He knows that Numi usually does the cooking. He asks:
	N:	nǔmí		nɛ́	á	kʉ́?
		Numi		cook	foc	what
		‘What has Numi cooked?’
	Ng:	nǔmí		nɛ́	á	kǝ̀ló	# à	nɛ́	á	mbà	bǝ̀loŋ
		Numi		cook	foc	plantains	she	cook	foc	also	potatoes
		intended: ‘Numi has cooked PLANTAINS. She has also cooked POTATOES’.

iii. EXH-driven entailments (Hartmann and Zimmermann 2007a): This test checks for the presence of exhaustivity inferences in a more indirect way. It tests for the availability of a further deduction, the validity of which depends on an exhaustive interpretation of the á-marked focus constituent. In (51), Watat wants to find out whether Nana invited him. Here, Ngami’s answer that the á-focused Numi was not invited is sufficient for Watat to safely conclude that he was. In other words, the basis for Watat’s deduction is the exhaustive focus information that Numi and nobody else was not invited.

(51)	Context:		Watat wants to know whether he is among the persons that Nana invited. His friend Ngami knows that he is but he had promised Nana not to tell Watat about whether he is invited or not. This promise does not bind Ngami to talk about other people, though. So he says:
	mbʉ́		mǝ̀	kʉ̀	sɔ̀ŋ	ú	yú	ndà’ndʒʉ̀ but		mbʉ́
	can		1sg	neg	tell	2sg	yours			can
	mǝ̀		tʃúb	ú	mbʉ̀	nnά	kʉ̀	síàŋ	á	nǔmí
	1sg		tell	2sg	that	Nana	neg	invite	foc	numi
	‘I cannot tell you about yourself, but I can tell you that Nana did not invite NUMI’.

When questioned whether Watat can safely deduce that he was or was not invited, the five speakers consulted answered that ‘Watat can easily deduce that he is invited, since the only person who is not invited is Numi’. Again, this directly follows on from an inherent exhaustive interpretation of the á-marked in situ focus. Note in passing that the same deduction does not automatically go through with plain accent focus in English, which is semantically non-exhaustive.

iv. And/but- continuations: The final EXH- test is taken from Beaver and Onea (2009) and can be considered a variant of the tests with additive follow-ups. The logic behind this test is that inherently exhaustive focus constructions are incompatible with an additive follow-up, and that any such follow-ups would need to be modulated with the discourse-contrastive marker ‘but’. When presented with the two variants in (52), the but- follow up in (52b) was preferred over the plain and- continuation in (52a). The preference for contrast-marking but is induced by the exhaustive interpretation of á-marked focus in the first clause, which must be reconciled with the continuation through the use of the contrastive discourse marker but.12

(52)	a.	# nǔmí	ʒʉ́n	á	túnzwə́	ŋ̌	á	bɛ̀n	ʒʉ́n	mbà	tʃʉ̀ʔ
		Numi	buy	foc	shirt	yes	3sg	again	buy	also	hat
		‘Numi bought á shirt. Yes, and he also bought hat’.
	b.	nǔmí	ʒʉ́n	á	túnzwə́	ŋ̌	ndàʔndʒʉ́	á	ʒʉ́n	mbà	tʃʉ̀ʔ
		Numi	buy	foc	shirt	yes	but	3sg	also	also	hat
		‘Numi bought á shirt. Yes, but he also bought hat’.

In summary, this subsection provides a wealth of evidence for the semantically exhaustive nature of á-marked in situ foci. We can conclude that the need to express an exhaustive inference cannot be the reason behind A’ focus movement in Mǝ̀dʉ́mbà. Since corresponding in situ information foci without á are non-exhaustive, it also follows that the EXH- inference is triggered as part of the lexical meaning of the contrastive focus marker á, which needs to be modified accordingly. We will provide a formal meaning of á in Section 4.1.3.

4.1.2. A Comment on Keupdjio’s (2020) Tests against EXH- Inferences with á

Contrasting with our findings from the previous subsection, Keupdjio (2020) puts forward some putative evidence against the inherent EXH- nature of á, which we will quickly address and set aside here. Adopting a test from Szabolsci (1981) and Kiss (1998), he claims on p.70 that the á-marked conjunctive answer in (53A1) would entail the individual á-marked conjuncts with á-bɔ̀ and á-tʃʉ̀^ʔ, respectively. This judgment is not shared by the second author and seven other speakers consulted by us. In particular, all of the other speakers rejected the somewhat more natural variant of his example in (53A2) as contradictory, specifying that the focus marker is infelicitous when replying to the person who answered in A1.13 Crucially, though, (53A2) should be fine if there were no EXH- entailment.

(53)	Q.	Nuga	náʔ	fά	á	kʉ́	Wàtɛ̀t	á
		Nuga	pst.rem	give	foc	what	Watat	qm
		‘What did Nuga give to Watat?’
	A1.	à	náʔ	fά	á	bɔ̀	bǔ	tʃʉ̀ʔ	yí
		3sg	pst.rem	give	foc	bag	and	hat	3sg
		‘He gave a BAG AND A HAT to him’.
	A2.	# ŋ̌	à	náʔ	fά	á	bɔ̀	yí.	nsǝ̀’ǝ̀	mbà
		yes	3sg	pst.rem	give	foc	bag	3sg	come	even
		tʃʉ̀ʔ	lá	mǝ̀	kʉ̀	lɛ̀n
		hat	def	1sg	neg	know
		‘Yes, he gave a BAG to him. I don’t know about the hat, though’.

Keupdjio’s (2020, p. 71) second argument concerns the fact that the á-marked focus constituent can co-occur with a universal quantifier. As first argued in Kiss (1998), this should not be possible if the focus marker comes with an EXH- inference. Keupdjio (2020) presents the following grammatical example, but unfortunately again without context:

(54)	Nuga	náʔ	fά	á	ndʒɔ̌ŋ	tʃʉ̀ʔ	fɛ́	Wàtɛ̀t	(Keupdjio 2020: 71, ex.100)
	Nuga	pst.rem	give	foc	every	hat	all	Watat
	‘Nuga gave every hat to Watat’.

Crucially, though, (54) cannot be used as an answer to the object question ‘What did Nuga gave to Watat?’, which would involve universal quantification over the domain of individual hats, thereby rendering exhaustive quantification vacuous. Instead, (54) can only be used in contrastive statements with narrow focus on the universal quantifier (or the polarity). For instance, it can correct the negated statement in (55). When used in this context, (54) felicitously expresses exhaustive and contrastive focus on the quantifier every, as opposed to not every.

(55)	Nuga	náʔ	kʉ̀	fά	á	ndʒɔ̌ŋ	tʃʉ̀ʔ	fɛ́	Wàtɛ̀t
	Nuga	pst.rem	neg	give	foc	every	hat	all	Watat
	‘Nuga didn’t give all the hats to Watat’

We therefore conclude that (54) is not a valid counterexample to the claim that á comes with an EXH- inference.

A third argument comes from the co-occurrence of the in situ focus marker á with the additive particle mbà ‘also, too, even’, which should not be licit if á were exhaustive. Keupdjio (2020, p. 72) presents the example in (56) as felicitous in support of his non-EXH analysis. Again, the example is presented out of context and, again, we were unable to replicate this judgment with our Medumba consultants, at least on an additive reading.14

(56)	Nuga	náʔ	fά	á	mbà	tʃʉ̀ʔ	Wàtɛ̀t	(Keupdjio 2020, p. 72, ex.101, his paraphrase)
	Nuga	pst.rem	give	foc	even	hat	Watat
	‘Nuga gave [even the hat]FOC to Watat’.

In fact, closer scrutiny shows that (56) may be marginally felicitous, but only on an exhaustive mirative interpretation driven by the scalar even-meaning of mbà (as already evident in Keupdjio’s telling paraphrase). In this construal, (56) can mean that Nuga rather gave a hat to Watat when Watat was expecting something else. Crucially, the sentence does not say that Nuga gave anything else to Watat, in accordance with the presence of the focus marker á.

In sum, we have shown that Keupdjio’s arguments against an exhaustive interpretation of in situ focus with á are invalid. In part, we were unable to replicate some acceptability judgments, cf. (53). In other cases, the presentation of the data in Keupdjio (2020) was marred by the absence of context, cf. (54) and (56). Taken together, the data in (53) to (56) are consistent with our claim that the focus marker á triggers an EXH- inference as part of its lexical meaning. This, in turn, means that A’ focus fronting is not driven by the need to express an EXH- inference; see also Ylinärä et al. (2023) for similar findings for Finnish and Italian.

4.1.3. FM á Triggers a Not-at-Issue EXH- Inference

We conclude the discussion of the exhaustive nature of á by showing that its EXH- inference is not at issue. The original test goes back to Szabolcsi (1994) for Hungarian, but more recently it has also been applied to the focus marker lɛ in Awing, Mǝ̀dʉ́mbà’s Grassfields Bantu cousin (Fominyam and Šimík 2017). The following contrast between (57a) and (57b) shows that exhaustiveness with á marking is invisible to negation and, hence, not truth-conditional, unlike what is found with the truth-conditional exclusive particle ndà^ʔ ‘only’. This can be seen from the fact that the additive particle is infelicitous in (57a) with á-marked focus, whereas it is required for felicity in (57b) with the exclusive, which explicitly negates the at-issue exhaustivity claim on bananas.

(57)	a.	nǔmí	kʉ̀	bʉ̀	swɛ̀n	á	bànànà	ndàʔndʒʉ̀	á	(#mbà)	pùmá
		Numi	neg	hab	sell	foc	banana,	but	foc	also	orange
		‘Numi does not sell BANANAS, but (#also) ORANGES’.
	b.	nǔmí	kʉ̀	bʉ̀	swɛ̀n	á	ndàʔ	bànànà	ndàʔndʒʉ̀	#(mba)	pùmá
		Numi	neg	hab	sell	foc	only	banana,	but	also	orange
		‘Numi does not only sell BANANAS, but also oranges’.

The second test goes back to Horn (1981) and shows the same thing. The EXH- inference of in situ á differs from the truth-conditional at-issue entailment of the exclusive particle ndà^ʔ. In particular, the at-issue EXH- entailment of ndà^ʔ makes the follow-up clause in (58b) informative, as it adds information to the proposition of the first clause. In contrast, the follow-up in (58a) is redundant, with or without á. This shows that the EXH- inference of á does not register at issue.

(58)	a.	# mǝ̀	lɛ̀n	mbʉ̀	nǔmí	bʉ̀	swɛ̀n	bànànà	ndà’ndʒʉ̀	à
		1sg	know	that	Numi	hab	sell	banana	but	3sg
		kʉ́	swɛ̀n	(á)	bànànà
		prog	sell	foc	banana
		intended: ‘I know that Numi sells banana but he sells BANANAS’.
	b.	mǝ̀	lɛ̀n	mbʉ̀	nǔmí	bʉ̀	swɛ̀n	bànànà	ndà’ndʒʉ̀	à
		1sg	know	that	Numi	hab	sell	banana	but	3sg
		kʉ́	swɛ̀n	á	ndàʔ	bànànà
		prog	sell	foc	only	banana
		‘I know that Numi sells banana but he sells only BANANAS’.

In light of this, we propose the following lexical entry for the exhaustive contrastive focus marker á in Mǝ̀dʉ́mbà. Á takes an individual, a backgrounded predicate, and a set of alternatives as its complements. It leaves the at-issue meaning component unchanged, and it adds a not-at-issue inference to the effect that the backgrounded predication does not hold for any of the contextually salient alternatives:15

(59)	Lexical meaning of focus marker á:
	〚 á 〛 = λx<σ>.λP<σ,t>.λC<σ,t>. P(x)	● for all z∈C: ¬P(z)     not at-issue

4.2. Focus Movement Triggers an Existence Presupposition

In this subsection, we show that focus movement triggers an existence presupposition on the backgrounded predication. The obligatory existence of existence presuppositions has been observed for English focus it- clefts (Horn 1981) and for syntactic focus fronting in the Kwa languages Ga and Akan (Grubic et al. 2019). The same holds for Mǝ̀dʉ́mbà focus fronting.

We show this by applying two standard tests for existence presuppositions: (i.) the (im)possibility of focused negative quantifiers or negated indefinites, and (ii.) the (in)felicity of focus-marked constructions in contexts in which the speaker does not commit to the existence presupposition (Rooth 1996). See Grubic et al. (2019, Section 6) for an extended discussion of these diagnostics applied to other West African languages.

As for the first, consider the four possible answers to the question ‘What did Nana cook?’ in (60A1–A4): (A1) in situ focus without á, (A2) in situ focus with á, (A3) ex situ focus fronting, and (A4) ex situ focus clefting with relative substructure. When the answer is negative (Nana cooked nothing), the in situ foci in (60A1) or (60A2) without focus fronting and focus clefting are the only felicitous options. Once again, the infelicity of focus fronting of smʒú in (60A3, A4) was confirmed by all four of the other native speakers consulted.

(60)	Q1.	What did Nana cook?
	A1.	nnά	nɛ́	smʒú
		Nana	cook	nothing
	A2.	nnά	nɛ́	á	smʒú
		Nana	cook	foc	nothing
		‘Nana cooked nothing’
	A3.	#á	smʒú	nnά	nɛ́	lá
		foc	nothing	Nana	cook	C.DEF
	A4.	#á	smʒú	[zǝ̀	nnά	nɛ́	lá]
		foc	nothing	rel	Nana	cook	c.def
		#‘It is nothing that Nana cooked’.

The contrast between (60A2) and (60A3) shows clearly that it is not the focus marker á, but the syntactic fronting operation that triggers an existence presupposition, which clashes with the at-issue content that nothing was cooked.

The data in (61A1–A3) show the same. Here, we embed the same four focus structures in a context in which the speaker is not committed to the existence of an individual satisfying the backgrounded predicate (Rooth 1996). If the focus construction in question triggers an existence presupposition, this will lead to a conflict with the speaker’s openly expressed non-commitment, thereby leading to a contradiction.

(61)

context:

We know that Watat’s brother Nana does not usually cook, but if he does, he only cooks plantain, never anything else. His friend Numi visits the family and asks Watat whether Nana cooked anything. Watat answers:

A1.

mǝ̀

kʉ̀

kwɛ́dǝ̀

nùmbʉ̀

mǝ̀

kʉ̀

kwɛ́dǝ̀

bʉ̀

[ à

nɛ́

(á)

1sg

neg

think

because

1sg

neg

think

C

3sg

cook

foc

kǝ̀lo ]

…

[(á) in situ]

plantain

A2.

#mǝ̀

kʉ̀

kwɛ́dǝ̀

nùmbʉ̀

mǝ̀

kʉ̀

kwɛ́dǝ̀

bʉ̀

[ á

kǝ̀lo

à

1sg

neg

think

because

1sg

neg

think

C

foc

plantain

3sg

nɛ́

lá ]

…

[á ex situ]

cook

c.def

A3.

#mǝ̀

kʉ̀

kwɛ́dǝ̀

nùmbʉ̀

mǝ̀

kʉ̀

kwɛ́dǝ̀

bʉ̀

[ á

kǝ̀lo

zǝ̀

1sg

neg

think

because

1sg

neg

think

C

foc

plantain

rel

à

nɛ́

lá ]

…

3sg

cook

c.def

…

à

kʉ̀

bʉ̀

nɛ́

ʒú

vwà

3SG

neg

hab

cook

something

else

‘I don’t think so because I don’t think he cooked PLANTAIN, and he didn’t cook anything else’.

Drawing on the judgments of the second author and four other native speakers, the crucial difference in felicity is between the felicitous in situ focus variant with or without á in (61A1), on the one hand, and the two infelicitous focus-initial constructions in (61A2) and (61A3) on the other. All speakers consulted agreed that (61A2) with focus movement and its cleft variant in (61A3) are contradictory. We take the infelicity of (61A2) to show that focus A’ movement triggers an existence presupposition, the same as the overt cleft variant in (61A3).

Summing up, this section has shown that A’ focus movement in Mǝ̀dʉ́mbà does have a discourse-semantic effect in the form of an existence presupposition. Mǝ̀dʉ́mbà focus movement thereby resembles A’ focus movement in other West African languages, such as Akan and Ga (Grubic et al. 2019).

4.3. Remaining Questions

Concluding this section, we will briefly address two other issues: First, we do not want to rule out the possibility that syntactic A’ movement of the focus constituent to a sentence-initial position serves other, softer discourse-pragmatic functions in addition to introducing the existence presupposition. After all, it is well known that the sentence-initial position is particularly prominent from a communicative perspective. As such, it should not come as a surprise if it were exploited for other discourse-pragmatic functions, such as directing the addressee’s attention to the focus denotation. In the functionalist tradition, this was captured in the form of Gundel’s (1988) First Things First Principle or Givón’s (1988) principle (Attend to the most urgent task first). More recently, Hartmann and Zimmermann (2007c) and Zimmermann (2008) have linked syntactic focus movement in Hausa and other Chadic languages to the notion of emphasis, where emphasis is understood in terms of unexpectedness of the focus denotation for the addressee and/or speaker in the given context. If the focus denotation is unexpected for the speaker, this effect is also referred to as mirativity (DeLancey 1997; Peterson 2016). In two experimental studies, for both production and acceptability, Destruel and Velleman (2014) showed that focus it- clefts in English are produced more often, and rated better, in contexts of discourse unexpectedness. A typical context instantiating strong discourse unexpectedness on the side of the addressee is a context with counter-presuppositional focus, already discussed in connection with (41) above, in which the speaker corrects a false background assumption of the addressee. Example (62) from Destruel and Velleman (2014) illustrates a context where the it-cleft correction received higher ratings:

(62)	Speaker A:	This bean dip is fantastic. I really want to get the recipe …
		… I can’t believe that Shannon brought it—she’s normally not a very good cook.
	Speaker B:	It was Tim who made it.

The greater acceptability and production rate of it- clefts with unexpected foci suggest that focus fronting can be exploited for additional pragmatic functions next to its basic discourse-semantic contribution, the existence presupposition. Further, Tönnies (2021) argues that focus it- clefts in German and English serve a specific discourse-structuring function, which is also related to discourse unexpectedness (and may even subsume (62) above as a special case). In particular, she argues that it- clefts address a question that is not the most relevant immediate question under discussion at the current state of discourse. Example (63) illustrates an example where the it- cleft (63a) is preferred over the canonical sentence structure in (63b). Crucially, at the time of utterance, the question about the identity of Lena’s new acquaintance has shifted somewhat into the background.

(63)	Context:		Yesterday at the party, Lena talked to some guy1. The two of them laughed a lot and they agreed to meet again the next evening. Then, Lena went home happily.
		a.	It was Peter that she talked to.
		b.	?She talked to Peter.

Given these observations for focus cleft structures in European languages, we think it likely that focus movement in Mǝ̀dʉ́mbà may likewise be exploited for additional pragmatic effects, such as the expression of interlocutor attitudes and discourse structuring. This claim would require further investigation in the form of corpus studies and controlled discourse-semantic experimenting, however—a challenging task that we will set aside for another occasion.

Finally, let us briefly return to the question of what blocks A’ focus fronting with temporal, spatial, instrumental, or causal adjuncts. This was illustrated in (12), repeated here for convenience:

(12)	* á	ŋgɣuʔ-muʔ	Nuŋgɛ̀	náʔ	nʤʉ̂n	Watɛ̀t	lá
	foc	year-other	Nuga	pst.rem	see	Watat	c.def
	INTENDED: ‘It was last year that Nuga saw Watat’.

In the work of Keupdjio (2020), the impossibility of fronting such adjuncts was attributed to the putative exhaustifying function of focus movement, which he argued would be incompatible with non-individual-denoting expressions, such as function-denoting adjuncts of type <v,t> (v the type of events). However, given that such focused adjuncts are felicitous in situ not only in the presence of the exhaustifying focus marker á (64a), but even with the at-issue exclusive particle ndà^ʔ ‘only’ (64b), we contend that this explanation cannot be correct.

(64)	a.	Nuŋgɛ̀	náʔ	nʤʉ̂n	Watɛ̀t	á	ŋgɣuʔ-muʔ
		Nuga	pst.rem	see	Watat	foc	year-other
		‘Nuga saw Watat last year’.
	b.	Nuŋgɛ̀	náʔ	nʤʉ̂n	Watɛ̀t	á	ndàʔ	ŋgɣuʔ-muʔ
		Nuga	pst.rem	see	Watat	foc	only	year-other
		‘Nuga saw Watat only last year’.

Moreover, focused adjuncts can also be focus-clefted, as shown in (65). This shows again that the reason for the ungrammaticality of (12) cannot be semantic in nature.

(65)	à	bʉ́	á	dàmnʒʉ́	[zǝ̀	nnά	àʔ	nɛ́	bάn	lá ]
	3sg	cop	foc	tomorrow	rel	nana	fut	cook	fufu	c.def
	‘It is tomorrow that Nana will cook fufu’.

At this point, we do not have a conclusive answer as to why focus movement with adjuncts, as in (12), is systematically ruled out in Mǝ̀dʉ́mbà. One explanation could be that focus fronting is blocked with framing constituents, possibly supporting claims by Frascarelli (2017) that focus and frames are distinct IS- categories with different (syntactic) properties. Alternatively, the reason for the ill-formedness of (12) could simply be syntactic in nature, and not semantic. Since the structure in (12) involves A’ movement, as plausibly argued in Keupdjio (2020), it seems as if the movement of adjuncts is illicit if it crosses a syntactic boundary, cf. (68a), as opposed to proper arguments, cf. (66b).16 Note that the presence of such an additional clause-internal CP/DP layer may actually be indicated by the clause-final definite determiner lá. In contrast, the overt cleft structure in (65) seems to involve a bipartite structure with covert operator movement inside the backgrounded relative clause; that is, without crossing the additional DP boundary, cf. (67).

(66)	a.	*ADJFOC [CP/DP  SVO tADJ lá ]
	b.	OBJFOC [CP/DP  SV tOBJ lá ]
(67)		à bʉ́  ADJFOC,i [DP [CP OPi SVO tADJ ] lá ]

We leave a more thorough syntactic investigation to another occasion.

5. Conclusions

In this paper, we gave an in-depth description and the first formal analysis of the realisation of focus in the Grassfields Bantu language Mǝ̀dʉ́mbà, a language with consistent morphological focus marking and optional A’ focus fronting. We have shown that two semantic factors that are often proposed as triggers for grammatically optional focus marking, namely, contrast and exhaustivity, are systematically expressed as part of the lexical meaning of the focus marker á, even with in situ focus. It follows that contrast and exhaustivity cannot be the driving forces behind optional focus movement, in Mǝ̀dʉ́mbà at least; see also Ylinärä et al. (2023). Thus, there seems to be a systematic difference between morphological focus-marking languages, such as Mǝ̀dʉ́mbà, on the one hand, and focus-accenting languages, such as Russian, on the other: focus-accenting languages such as Russian may have to resort to focus fronting for the expression of contrastive focus in the absence of a categorical indication of contrast in the acoustic signal (Titov 2012), whereas morphological focus-marking languages can mark contrast in situ. We have also shown that focus movement comes with an additional semantic effect in the form of an existence presupposition. The triggering of existence presuppositions is familiar from syntactic fronting in other West African languages (Grubic et al. 2019), and it is also a typical feature of focus–background partitions in cleft sentences. Finally, we speculate that the non-canonical fronting structures in Mǝ̀dʉ́mbà may be exploited to achieve other, softer discourse-pragmatic effects, such as guiding the attention of the addressee, or to highlight the focus denotation as discourse-unexpected.