1. Introduction

1Safir et al. (2020) demonstrate the existence of the so-called “actual clause” in Lubukusu (JE31c). The nature of this clause is understood by contrasting it with matrix indicative clauses and clausal complementation contexts. In a matrix sentence, if the speaker uses the past tense as in Nafula ate the cake, then the speaker is committed to the truth of the statement because it is an assertion. We take the act of making an assertion to mean that a proposition is introduced by the speaker, that the speaker is committed to the truth of the proposition, and normally that the proposition is new information. If we take a sentence like Wekesa says/believes that Nafula ate the cake, the speaker’s assertion is that Wekesa has a particular belief, but the speaker has no commitment to the truth of Nafula ate the cake. What is striking about actual clauses is that when the actual is used in the says/believes complement, then the speaker is also committed to the truth of Nafula ate the cake. In Safir et al. (2020), this property was treated as an actuality entailment, much in the way that the statement John managed to eat the cake entails that John ate the cake (with whatever effort it required). In this paper we take a different tack, treating actual clauses as embedded assertions whereby the speaker is committed to the truth of the complement clause proposition.

2In languages like English, the speaker’s commitment to the truth of the embedded proposition can be achieved by adding a special adverb or a continuation attached to the sentence, as shown in parentheses in (1). In addition, there are certain verbs, like manage, that entail the truth of their complement proposition. For examples like (2), the parenthetical sounds redundant.

(1)	Aya told Mary to eat the fish (and she did eat the fish).

(2)	Aya managed to eat the fish (#and she did eat the fish).

3In English, there is no special clause type that encodes the parentheticals in (1) and (2), but in many Bantu languages there is distinct morphology in the embedded clause, bolded in (3), which adds the force of the parenthetical. We gloss the morpheme in question as act (for actual). The parenthetical clause in translation is used to make the reading clear, but the meaning in question is denoted only by the presence of ‑ka‑ in the subordinate clause.

(3)	Aya a‑li‑mw‑amb‑ia	Marya	a‑ka‑m‑la	samaki
	Aya sm.c1‑pst‑om.c1‑tell‑fv	c1.Mary	sm.c1‑act‑om.c1‑eat	c9.fish
	‘Aya told Mary to eat fish (and she did eat fish).’ [Kiswahili]

5We will proceed as follows. In Section 2 we will review the distribution of the actual clause complement and describe the effects that actual clauses evidence. Section 3 analyzes the properties of actual clause complements from the embedded-assertion point of view. In Section 4, we briefly show that what we are calling “actual morphology” in Bantu more generally is also used to signal clause-chaining in narratives. We introduce our contingent antecedent clause analysis, and we show that our account predicts that the actual clause complement effect follows from the same semantic function that actual morphology signals in clause-chaining contexts. In our conclusion, we discuss what typological implications our proposal might have across Bantu, and perhaps more generally.

2. Identifying actual clauses as complements

6The embedded clause in (3) is what we characterize as an actual clause. Note that what we are calling “actual morphology” has been described variously in Bantu and other languages as a narrative/consecutive/sequential tense/aspect marker (Dahl 1977 and 1985; Johnson 1979; Schröder 2013; Englebretson & Wambũi 2015). The actual clause distribution diverges from what is generally expected of the same morpheme used in narrative contexts. In particular, the actual occurs as a complement of non-factive verbs, while the narrative use seems to be a case of clause-chaining. We shall have more to say about the relation between the actual complement effect and narrative clauses in Section 4. Nevertheless, we show that the actual clause complement effect, though not discussed in the literature outside of Lubukusu, has a larger presence in the Bantu family of languages.

8There are various arguments to treat actual clauses as complement clauses (not adjuncts). First, actual clauses are tenseless and are dependent on the tense of the matrix clause for time reference. Second, actual complement clauses can be compatible with verbs that do not take indicative clauses as complements (Safir et al. 2020), but that do take subjunctive and/or infinitive complements. Notice that the applicative verbs in (4a) and (4b), which normally require two objects, have the actual clause as a complement that counts as one of the objects. In Kiswahili (4b), where the om is generally agreed to be an agreement marker (e.g. Keach 1995), the presence of the om does not affect the verb’s argument structure, and so the actual clause remains a complement. Without the actual clause, the translation would be #’Juma/Obuchi tried for the child’.

9Additional evidence that the actual clause is a complement is that it allows extraction, as illustrated for Lubukusu and Kiswahili in (5).

(5)	a.	Y‑á‑b‑a	e‑ng’eni	níyo	Wekesa
		sm.c9‑pst‑be‑fv	c9‑fish	comp.c9	Wekesa
		á‑subila/é‑nya/á‑loma		Mary	a‑a‑ly‑a.
		sm.c1.pst‑believe/want/say		Mary	sm.c1‑act‑eat‑fv
		‘It was the fish that Wekesa believed/wanted/said Mary act‑ate.’ [Lubukusu]
	b.	Ilikuwa samaki ndiyo Wekesa aliamini/taka/sema Marya akaila.
		I‑li‑kuw‑a	samaki	ndiye	Wekesa
		sm.c9‑pst‑be‑fv	c9.fish	comp.c1	Wekesa
		a‑li‑amini/taka/sema		Marya	a‑ka‑m‑l‑a
		sm.c1.pst‑believe/want/say		Mary	sm.c1‑act‑om.c1‑eat‑fv
		‘It was the fish that Wekesa believed/wanted/said Mary act‑ate.’ [Kiswahili]

10The typical assumption is movement is only possible if the clause is a complement, and that adjuncts are islands for movement. Although there are probably certain kinds of low adjuncts without overt subjects that can be extracted from in English (e.g. Which bill did John leave town without paying?), sometimes called converbs (see Haspelmath 1995: 10), extraction from actual clause contexts does not fit Haspelmath’s converb profile, whichmost typically lack an overt subject.

11However, there is a result interpretation that arises with some predicates that is somewhat unlike typical complementation relations. The matrix verb event seems to have a causal relation to the truth of the prejacent proposition in most instances (see also fn.1). Thus (4) is most naturally accepted if as a result of Juma’s efforts, the child did go. Even for verbs meaning ‘believe’, which can typically take actual clause complements, the result implication can also be present (at least for Lubukusu and Kiswahili). We return to this point later. As we establish, however, certain semantic properties of the matrix predicate must be compatible with actual clause entailments.

12Essentially, the actual clause complement describes events that must have happened according to the utterer of the sentence (henceforth, UTT, as distinguished from a reported speaker). In several languages like English, this is expressed by the past tense in a matrix clause, but whether or not past in a subordinate clause describes an event that must have happened depends on what sort of matrix verb is used. When the actual is used in a complement, then the complement must also be something that UTT asserts to have actually happened or to be the case. This is shown in the contrast between the pairs of sentences for Lubukusu, Kiswahili and Luganda in (6)‑(8), respectively. In response to a reviewer, we show (with a Luganda example) that the presence of an indirect object in the matrix clause does not change the influence of actual morphology. Examples for other languages in our sample (see Table 1) include just the actual clause version to save space, as in (9).

(6)	a.	Wekesa á‑ból‑el‑a		o‑mw‑aana	a‑li	Nafula
		Wekesa sm.c1.pst‑tell‑appl‑fv		c1‑c1‑child	c1‑comp	Nafula
		á‑ly‑á	e‑keki.
		sm.c1.pst‑eat‑fv	c9‑cake
		‘Wekesa told the child that Nafula ate the cake.’ [Lubukusu]
	b.	Wekesa á‑ból‑el‑a		Nafula a‑a‑ly‑a		e‑keki.
		Wekesa sm.c1.pst‑tell‑fv		Nafula sm.c1‑act‑eat‑fv		c9‑cake
		‘Wekesa told Nafula to eat the cake (and she did eat the cake).’ [Lubukusu]

(7)	a.	Peter alimwambia mtoto kuwa Mary alikula keki.
		Peter	a‑li‑mwamb‑i‑a		mtoto	kuwa
		Peter	sm.c1.pst‑tell‑appl‑fv		c1‑child	comp
		Maria	a‑li‑kul‑a		keki
		Maria	sm.c1.pst‑eat‑fv		c9.cake
		‘Peter told the child that Mary ate the cake.’ [Kiswahili]
	b.	Aya	a‑li‑mw‑amb‑ia	Marya	a‑ka‑m‑la		samaki.
		Aya	sm.c1‑pst‑om.c1‑tell‑fv	Mary	sm.c1‑act‑om.c1‑eat		c9.fish
		‘Aya told Mary to eat fish (and she did eat fish).’ [Kiswahili]

(9)	a.	Aya	a‑ka‑mo‑teb‑ia	Maria	a‑ka‑ri‑a	enswe
		Aya	sm.c1‑pst‑om.c1‑tell‑fv	Mary	sm.c1‑act‑eat‑fv	c9.fish
		‘Aya told Mary to eat fish (and she did eat fish).’ [Ekegusii]
	b.	Aya	wa‑ka‑mw‑ambir‑a	Marya	a‑chi‑ri‑a	samaki.
		Aya	sm.c1‑pst‑om.c1‑tell‑fv	Mary	sm.c1‑act‑eat‑fv	c9.fish
		‘Aya told Mary to eat fish (and she did eat fish).’ [Kidigo]
	c.	Aya	a‑á‑bwir‑ye	Mariya	a‑ra‑ri‑a	isamaki.
		Aya	sm.c1‑pst‑tell‑fv	Mary	sm.c1‑act‑eat‑fv	c9.fish
		‘Aya told Mary to eat fish (and she did eat fish).’ [Kirundi]

13Notice that the complement clauses in the (a) examples of (6)‑(8) have a past morpheme that is morphologically different from the morphology in the complement clauses in (b). In Lubukusu, the past uses a high-toned short vowel which also doubles as the class 1 subject marker, while the actual is marked by a long toneless vowel. In Luganda, both the actual and the past are a short vowel ‑a‑, but the former is distinguished by what looks like a dependent marker n‑ which precedes the actual marker. The distinctive actual morphology is illustrated in (9a‑c) for other languages in our sample. Notice also that the actual clause complements typically lack a complementizer.

15Actual clauses in all the languages we discuss are interpreted as making a statement about events or states known to be true by UTT at the time of the utterance and not those that have not happened. This is confirmed by a negation test applied to the complement clauses. It is expected that if a subordinate clause expresses what is believed to have actually happened in the real world, then any contradiction of the subordinate proposition results in unacceptability. In the Kiswahili, Kidigo, Kikamba, Lubukusu, and Kirundi examples in (10a‑e), the follow-up negation in brackets is infelicitous because in the world of UTT, Wafula did go, and hence the truth cannot be negated. Similar results are also reported for Luganda by our consultants, but we do not have an example.

(10)	a.	Wekesa	a‑li‑m‑himiza	Wafula	a‑ka‑end‑a
		Wekesa	sm.c1‑pst‑om.c1‑urge	Wafula	sm.c1‑act‑go‑fv
		[*lakani	ha‑a‑ku‑end‑a].
		but	neg‑pst‑c15‑go‑fv
		‘Wekesa urged Wafula to go (and he did) [*and he did not go].’ [Kiswahili]
	b.	Wekesa	wa‑mw‑imiz‑a	Wafula	a‑chi‑phiy‑a
		Wekesa	sm.c1.pst‑om.c1‑urge‑fv	Wafula	sm.c1‑act‑go‑fv
		[*ila	ka‑ya‑phiy‑a].
		[but	neg‑sm.c1.pst‑go‑fv]
		‘Wekesa urged Wafula to go (and he did) [*but he did not go].’ [Kidigo]
	c.	Juma a‑mw‑isuv‑ie		Zalo	a‑nyw‑a	yiia
		Juma sm.c1.pst‑om.c1‑urge‑pfv		Zalo	sm.c1.act‑drink‑fv	milk
		[*indi	nda‑a‑nyw‑a].
		[but	neg‑sm.c1.pst‑drink‑fv]
		‘Juma urged Zalo to drink milk (and he did), [*but he did not drink].’ [Kikamba]
	d.	Wekesa	á‑subisy‑a	Wafula a‑a‑ch‑a
		Wekesa	sm.c1.pst‑urge‑fv	Wafula sm.c1‑act‑go‑fv
		[*nekakhali	sé‑á‑ch‑a	ta].
		but	neg‑sm.c1.pst‑go‑fv	neg
		‘Wekesa urged Wafula to go (and he did) [*and he did not go].’ [Lubukusu]
	e.	Wekesa	a‑á‑hanur‑ye	Wafula	a‑ra‑gend‑a
		Wekesa	sm.c1‑pst‑urge‑fv	Wafula	sm.c1‑act‑go‑fv
		[*ariko	nti‑a‑á‑gi‑ye].
		[*but	neg‑sm.c1‑pst‑go‑fv]
		Wekesa urged Wafula to go (and he did) [*and he did not go].’ [Kirundi]

17As Safir et al. (2020: 190) show for Lubukusu, the so-called actuality entailment does not originate in the matrix clause because the same matrix verb can take both subjunctive and infinitive complement clauses whose truth condition can be negated.

(11)	Wekesa		á‑a‑bólel‑a		Wafula	khu‑ch‑a
	Wekesa		sm.c1‑pst‑tell‑fv		Wafula	c15‑go‑fv
	(ne	kakhali	Wafula	se	á‑a‑ch‑a	ta).
	(and	though	Wafula	neg	sm.c1‑pst‑go‑fv	neg)
	‘Wekesa told Wafula to go, but Wafula did not go.’ [Lubukusu]

(12)	Wekesa	á‑a‑bólel‑a	Wafula	á‑ch‑e
	Wekesa	sm.c1‑pst‑tell‑fv	Wafula	sm.c1‑go‑sbjv
	(ne	kakhali Wafula	se	á‑a‑ch‑a	tá).
	(and	though Wafula	neg	sm.c1‑pst‑go‑fv	neg)
	‘Wekesa told Wafula that he may go, but Wafula did not go.’ [Lubukusu]

18In contrast to the actuality entailment introduced by the predicate manage in English, the effect of the actuality entailment is achieved in Lubukusu by the choice of complement.

20As Safir et al. (2020) have shown, UTT cannot use actual clause morphology to assert that a future event will take place, although matrix clauses with future tense or modal force are perfectly good assertions (but see Section 4.2). Because the actual clauses describe true events in the real world and because they are anchored to the tense of the matrix clause, they will always be incompatible with matrix events taking place after the present. This explains why verbs in future tense cannot take actual clauses even though future tense on those verbs is compatible with both infinitive and subjunctive, as shown for Lubukusu (14), Ekegusii (15) and Kiswahili (16).

(14)	a.	*Wafula	á‑khá‑eny‑e	a‑a‑bey‑a	Maria.
		Wafula	sm.c1‑fut‑want‑fv	sm.c1‑act‑marry‑fv	Mary
		intended: ‘Wafula will want that he married Mary (and he did).’ [Lubukusu]
	b.	Wafula	á‑khá‑eny‑e	khu‑bey‑a	Maria.
		Wafula	sm.c1‑fut‑want‑fv	c15‑marry‑fv	Mary
		‘Wafula will want to marry Mary.’ [Lubukusu]
	c.	Wafula	á‑khá‑eny‑e	a‑bey‑e	Maria.
		Wafula	sm.c1‑fut‑want‑fv	sm.c1‑marry‑sbjv	Mary
		‘Wafula will want that she may marry Mary.’ [Lubukusu]

(15)	a.	*Obuchi	n‑a‑g‑ani‑e	a‑ka‑nywom‑e	Maria.
		Obuchi	foc‑sm.c1‑fut‑want‑fv	sm.c1‑act‑marry‑fv	Mary
		intended: ‘Obuchi will want that he married Mary (and he did).’ [Ekegusii]
	b.	Obuchi	n‑a‑g‑ani‑e	ko‑nywom‑a	Maria.
		Obuchi	foc‑sm.c1‑fut‑want‑fv	c15‑marry‑fv	Mary
		‘Obuchi will want to marry Mary.’ [Ekegusii]
	c.	Obuchi	n‑a‑g‑ani‑e	a‑nywom‑e	Maria.
		Obuchi	foc‑sm.c1‑fut‑want‑fv	sm.c1‑marry‑sbjv	Mary
		‘Obuchi will want that he may marry Mary.’ [Ekegusii]

(16)	a.	*Wafula atataka akamwoa Maria.
		Wafula	a‑ta‑tak‑a	a‑ka‑mw‑o‑a	Maria
		Wafula	sm.c1‑fut‑want‑fv	sm.c1‑act‑om.c1‑marry‑fv	Mary
		intended: ‘Wafula will want that he married Mary (and he did).’ [Kiswahili]
	b.	Wafula atataka kumwoa Maria.
		Wafula	a‑ta‑tak‑a	ku‑mw‑o‑a	Maria
		Wafula	sm.c1‑fut‑want‑fv	c15‑om.c1‑marry‑fv	Mary
		‘Wafula will want to marry Mary.’ [Kiswahili]
	c.	Wafula atataka amwoe Maria.
		Wafula	a‑ta‑tak‑a	a‑mw‑o‑e	Maria
		Wafula	sm.c1‑fut‑want‑fv	sm.c1‑om.c1‑marry‑sbjv	Mary
		‘Wafula will want that he may marry Mary.’[Kiswahili]

21As Safir et al. (2020) point out for Lubukusu, verbs that take complements describing events that cannot take place cannot normally use the actual either, such as verbs meaning ‘fail’, ‘prevent’, or ‘deny permission’. The verb shindw‑ is not quite as bad as the other Kiswahili verbs in (17a).

22To summarize so far, we have established that the actual clauses in eastern Bantu, at least those in our dataset, have the following properties.

(18)	a.	Actual clauses have a distinctive morphology exponed by a dedicated morpheme, such as a vowel marked by tone or length, or a syllable like Kiswahili ‑ka‑. See Table 1.
	b.	The actual clause proposition denotes an event or state of affairs and is taken to be true by UTT.
	c.	The actual clause proposition denotes an event that has taken place or a state of affairs that holds at or before the time of utterance by UTT.
	d.	The actual clause proposition is a new contribution to the common ground.
	e.	An actual clause is only licensed as a complement clause.

23In what follows, we show how semantic restrictions signaled by actual morphology predict much of its distribution. In Section 4, we abandon (18e), as we extend our account to the narrative usage of actual morphology.

3. Actual clause interpretation in complement contexts

24We are proposing that the use of an actual clause as a complement necessarily treats the complement clause as an assertion, which entails that UTT is not only committed to the truth of the proposition in the assertion, but that UTT presents the proposition as new information to be added to the common ground.

25Recall that we mentioned that actual clauses cannot be complements to just any predicate that takes a complement clause. A striking fact is that actual clauses are infelicitous as complements to factive predicates, and we believe that our hypothesis that actual clauses function as assertions derives this pattern.

26In a paper on the selection of clausal types, Kiparsky & Kiparsky (1970) distinguish between factive and non-factive complements. A complement is said to be factive if it expresses a proposition taken to be true by UTT and the reported propositional attitude holder (the subject of the matrix verb), although there is more to say that we will get to later. The propositional attitude holder’s commitment to the truth of the complement proposition and that of UTT are presuppositions preserved under negation. Non-factive complements have no such presuppositions. In (19a), the use of the Lubukusu factive verb translated as ‘regret’ presupposes that Wekesa married Mary and that Alice also accepts this as a fact, even if the verb is negated, as in (20a). For the non-factive ‘believe’ (19b), Alice is committed to the truth of ‘Wekesa married Mary’, and UTT is not. As expected, negation removes Alice’s commitment to the truth of the complement clause proposition in (20b). The factive clauses require the use of the non-agreeing bali complementizer.

(19)	a.	Alice	á‑esóny‑a	bali	Wekesa	á‑a‑béy‑a
		Alice	sm.c1‑regret‑fv	comp	Wekesa	sm.c1‑pst‑marry‑fv
		Maria.
		Maria
		‘Alice regrets that Wekesa married Mary.’ [Lubukusu]
	b.	Alice	á‑subil‑a	a‑li	Wekesa	á‑a‑béy‑a
		Alice	sm.c1‑believe‑fv	c1‑comp	Wekesa	sm.c1‑pst‑marry‑fv
		Maria.
		Maria
		‘Alice believes that Wekesa married Mary.’ [Lubukusu]

(20)	a.	Alice	sé	á‑esóny‑a	bali	Wekesa	á‑a‑béy‑a
		Alice	neg	sm.c1‑regret‑fv	comp	Wekesa	sm.c1‑pst‑marry‑fv
		Maria	tá.
		Mary	neg
		‘Alice does not regret that Wekesa married Mary.’ [Lubukusu]
	b.	Alice	sé	á‑subil‑a	a‑li	Wekesa	á‑a‑béy‑a
		Alice	neg	sm.c1‑believe‑fv	c1‑comp	Wekesa	sm.c1‑pst‑marry‑fv
		Maria	ta.
		Maria	neg
		‘Alice does not believe that Wekesa married Mary.’

27Infinitive clauses are interpreted as denoting true propositions by both UTT and the propositional attitude holder when they occur as complements of factive verbs (21a, b), presuppositions that do not arise for non‑factive predicates like that in (21c), as indicated by the contradictions imposed by the parentheticals in (21a, b) but not (21c).

(21)	a.	Mary was surprised to have won the game, (#but she did not win it).
	b.	Mary was happy to have won the game, (#but she did not win it).
	c.	Mary planned to win the game, (but she did not win it).

28The Kiswahili example in (22a) shows that non‑factive verbs such as ‘plan’, ‘try’, and ‘believe’ can take actual clause complements while the factive ones like ‘regret’ in (22b) cannot. The question marks in the morpheme breakdowns are indications of gradations in the judgments and not part of the gloss proper.

(22)	a.	Ken	a‑li‑pang‑a/	a‑li‑jarib‑u/	[?]a‑li‑amin‑i/
		Ken	sm.c1‑pst‑plan‑fv/	sm.c1‑pst‑try‑fv/	sm.c1‑pst‑believe‑fv/
		[?]a‑li‑tak‑a		a‑ka‑end‑a	kwenye
		sm.c1‑pst‑want‑fv		sm.c1‑act‑go‑fv	to
		sherehe.
		c9.ceremony
		‘Ken planned/tried/believed/wanted to go to the ceremony (and he did go).’ [Kiswahili]
	b.	*Juma	a‑li‑sikitik‑a	a‑ka‑mw‑o‑a
		Juma	sm.c1‑pst‑regret‑fv	sm.c1‑act‑om.c1‑marry‑fv
		Marya.
		Mary
		intended: ‘Juma regretted that/and married Mary.’ [Kiswahili]

29The actual complement is also not possible for the factive interpretation when a verb allows both factive and non-factive readings (23).

(23)	Juma	a‑li‑kubali	Peter	a‑ka‑mw‑o‑a	Marya.
	Juma	sm.c1‑pst‑agree	Peter	sm.c1‑act‑om.c1‑marry‑fv	Mary
	* ‘Juma agreed that/and Peter married Mary.’
	‘Juma agreed for Peter to marry Mary, and he (Peter) did marry Mary’. [Kiswahili]

30The Kiswahili verb ‑kubali in (23) can be taken to be factive if it describes a situation where ‘that Peter married Mary’ is already accepted as true by the UTT and the addressee, in which case we might suppose that Juma has finally gotten around to the conclusion that UTT and addressee take to be correct. In the factive scenario, actual morphology in (23) is not accepted (indicated here by the * on the first translation). However, the verb ‑kubali can also be taken to indicate permission. That Juma agreed for Peter to marry Mary does not ensure that they were married, so the fact that they actually married subsequent to his agreement is what the actual can successfully assert.

31In Lubukusu, factive verbs that are incompatible with actual clauses include esony‑ ‘regret’, eyam‑ ‘accept’, funul‑ ‘reveal’, hukya‑ ‘surprise’, isindukh‑ ‘surprise’, ri‑ ‘fear’, and sangal‑ ‘rejoice’. A Lubukusu example using the verb fukilil‑ ‘admit’ shows that the same kind of distinction in Kiswahili (23) can also be illustrated for Lubukusu in (24).

(24)	Wekesa	á‑fukilil‑a	Peter	a‑a‑bey‑a	Marya.
	Wekesa	sm.c1.pst‑accept‑fv	Peter	sm.c1‑act‑marry‑fv	Mary
	‘Wekesa accepted for Peter to marry Mary (and he did).’ [Lubukusu]

33We acknowledge that our account of factivity is less subtle than more recent approaches that show that certain verbs can be non-factive in certain contexts without alternation of meaning of the verb (rather by manipulation of context) (see, e.g., Anand & Hacquard 2014, and references cited there). We have not had the opportunity to explore these subtleties crosslinguistically, but it is predicted that if a verb takes an actual complement, then it is interpreted as non-factive in the context where it is used. We hope to explore these questions in future research.

35To summarize this section, we have argued that complement actual clauses function as assertions, and as evidence we have demonstrated that factive predicates cannot have actual clause complements. We take UTT’s goal in making an assertion to be to present the addressee with new information that UTT expects the addressee to take as true. If UTT uses a factive predicate, then the truth of the factive complement is presupposed by UTT to be already accepted as true by the addressee. Thus using the actual clause as a subordinate assertion is incompatible with factive complementation.

4. Actual morphology in non-complement contexts

36We have noted that there seems to be a causal relation between the matrix verb clause and the complement clause, such that the truth of the complement clause seems to come about by virtue of the act of saying/believing/wanting etc., expressed by the matrix clause. Drawing on a part of an idea suggested by Leonard (1980), we will say that actual clauses are presented by the speaker (Leonard’s speaker is our UTT) as true because some contingency has been met. The “contingent clause antecedent” (our term) in the environment of complementation is the matrix clause, which, by virtue of being asserted true by the speaker, has the result of making the complement clause true. Actual clause morphology is what signals not only that the complement clause is true, but why it is true, that is, the contingency for making it true has been met.

37This way of putting things uses a small part of Leonard’s (1980) theory of the narrative tense in Kiswahili, which he took to signal that the event described by a Kiswahili clause with ka on the verb was to question the event. He used this description to cover three different cases. One case has to do with a rhetorical use where the proposition is taken by UTT to be true, but the appropriateness or surprise that the event/proposition is true is at stake. Without accounting for the rhetorical effect, we set this case aside because it does not challenge our assumption that the use of actual expresses an assertion (though there are other matters at stake to which we return). Another notion Leonard groups under “questioning the event” is to question whether the event is in fact true. This second notion is contradictory to ours, and so we will endeavor to reinterpret the rather slim evidence Leonard has for it later in this section. However, we would like to enlarge on the third notion Leonard groups under “questioning the event”, which is the idea that the event is contingent on some other proposition. What he does not note is that the contingency in question is taken to be met any time the narrative marker is used, so, in effect, the event is not questioned, but asserted to be true because its contingent antecedent proposition (our term) is accepted as true. This third notion is introduced by Leonard to account for the narrative use of ka whereby he takes meeting contingency to derive the consecutive effect in narrative as a consequence. Our revision of his idea follows his reasoning in this respect.

38Viewed through our elaboration of Leonard’s idea, the narrative use of ka treats the ka-marked proposition as the consequence of its contingent antecedent. Thus, after a scene-setting clause introduces a tense, each following ka-clause is a consequence of the last clause, which can be used to set up a sequence of events in a narrative. This is illustrated for Kiswahili, Ekegusii and Kirundi in (25). In Kiswahili, the clause following the adjunct clause is the one that sets the tense. The adjunct clause itself has an infinitive verb. The rest of the clauses have the narrative/actual morphology. This is also true for Lubukusu.

(25)	a.	Baada	ya	wa‑naume
		after	am.c9	c2‑man
		ku‑lim‑a	shamba,	wa‑nawake
		c15‑plough‑fv	c9.field	c2‑woman
		wa‑li‑pand‑a	mi‑mea,	wa‑ka‑nyunyizi‑a
		sm.c2‑pst‑plant‑fv	c4‑plant	c2‑act‑water‑fv
		mi‑che,	wa‑ka‑kusany‑a	ma‑zao,
		c4‑seedling	sm.c2‑act‑collect‑fv	c6‑crop
		kisha	wa‑naume	wa‑ka‑ingiz‑a
		then	c2‑man	sm.c2‑act‑take‑fv
		ndani	ma‑vuno.
		inside	c6‑harvest
		‘After the men plowed the field, the women planted the crops, watered the seedlings, staked the mature plants, and then the men brought in the harvest.’ [Kiswahili]
	b.	Nyuma	y‑a‑ba‑sacha	ko‑rem‑a
		after	am.c9‑c2‑c2‑man	c15‑plough‑fv
		o‑mo‑gondo,	a‑ba‑kungu	ba‑ga‑simek‑a
		c c3‑c3‑field	c2‑c2‑woman	sm.c2‑pst‑plant‑fv
		e‑bi‑meri,	ba‑ka‑bi‑imor‑er‑a
		c8‑c8‑plant	sm.c2‑act‑om.c2‑sprinkle‑appl‑fv
		a‑ma‑che,	ba‑ka‑sangerer‑i‑a	ri‑gesa,
		c6‑c6‑water	sm2‑act‑collect‑caus‑fv	c5‑plant
		erio	a‑ba‑sacha	ba‑ka‑ret‑a
		then	c2‑c2‑man	sm.c2‑act‑bring‑fv
		ri‑gesa.
		c5‑plant
		‘After the men plowed the field, the women planted the crops, watered the seedlings, staked the mature plants, and then the men brought in the harvest’. [Ekegusii]
	c.	A‑ba‑gabo	ba‑mar‑ye	ku‑rim‑a,
		c2‑c2‑men	sm.c1‑finish‑fv	inf‑cultivate‑fv
		a‑ba‑gore	ba‑ra‑ter‑ye	i‑bi‑terwa,
		c2‑c2‑women	sm.c2‑act‑plant‑fv	c8‑c8‑crops
		ba‑ra‑vomer‑a	i‑mbuto,	ba‑ra‑egerany‑a
		sm.c2‑act‑water‑fv	c9‑seedlings	sm.c1‑act‑collect‑fv
		i‑bi‑terwa	vyeze,	hanyuma
		c8‑c8‑plants	c8.mature	then
		a‑ba‑gabo	ba‑ra‑zan‑a	i‑vy‑imburwa.
		c2‑c2‑man	sm.c2‑act‑fv	c8‑c8‑harvest
		‘After the men plowed the field, the women planted the crops, watered the seedlings, staked the mature plants, and then the men brought in the harvest’ [Kirundi]

(26)	a.	Ná‑ech‑a,	na‑a‑bon‑a,	na‑a‑kuur‑a.
		sm.c1.pst‑come‑fv	sm.c1‑act‑see‑fv	sm.c1‑act‑conquer‑fv
		‘I came, I saw, I conquered.’ [Lubukusu]
	b.	Ni‑li‑kuj‑a,	ni‑ka‑on‑a,	ni‑ka‑shind‑a.
		sm.c1‑pst‑come‑fv	sm.c1‑act‑see‑fv	sm.c1‑act‑conquer‑fv
		‘I came, I saw, I conquered.’ [Kiswahili]
	c.	In‑ga‑ch‑a,	in‑ka‑ror‑a,	in‑ka‑bu‑a.
		sm.c1‑pst‑come‑fv	sm.c1‑act‑see‑fv	sm.c1‑act‑conquer‑fv
		‘I came, I saw, I conquered.’ [Ekegusii]
	d.	N‑á‑ra‑z‑ye,	n‑ra‑bon‑a,	n‑ra‑tsind‑a.
		sm.c1‑pst‑disj‑come‑fv	sm.c1‑act‑see‑fv	sm.c1‑act‑conquer‑fv
		‘I came, I saw, I conquered.’ [Kirundi]

40So far, we have only proposed that the actual morphology functions as a marker of an embedded assertion, but this does not suffice to connect the use of actual morphology in the complement clause to the use of actual morphology in narrative. In order to make this connection, we argue that both uses are licensed by dependency on a previous proposition in discourse. We will build our account of actual morphology licensing by refining what we take from previous accounts, and so we briefly explore those accounts in the next section.

4.1 Other accounts

41There are at least two distinct accounts of narrative constructions that will serve to distinguish what we are going to propose from what has previously been said.

42Our treatment of ACT morphology as marking an assertion differs notably from Leonard (1980) in a number of respects. In particular, Leonard’s second notion of questioning the event is directly incompatible with our theory. Specifically, he proposes that actual morphology (Kiswahili ka in his paper) signals that the truth of the actual clause proposition is in doubt. However, we think his defense of this view is based on a misinterpretation of a rhetorical use.

43Leonard defends his view by presenting the following passage drawn from an address by a leading historical figure, Julius Nyerere, here with Leonard’s gloss and translation.

(27)	Maskini wanaweza kuwa na roho za kibepari --- wanyonyaji wa binadamu wenzake.
	‘Poor men can have the souls of capitalists --- parasites on their fellow men.’
	Vile	vile,	Tajiri	anaweza	akawa na
	same	manner	rich	he.present.can	be.ka.be
	roho	ya	Ujamaa	Anaweza	akathamini
	with.spirit	of	socialism	he.present.can	he.ka.value
	mali	yake	kwa	sababu	tu
	wealth	his	for	reason	only
	inaweza	kutumiwa	kuwasaidia	binadamu	wenzake
	it.present.can	to.be.use.ps	to.them.help	human	companions.his
	Nimesema	kuwa	Tajiri	anaweza	akawa
	I.pfv.say	to.be	rich	he.present	he.ka.be
	mpenda	Ujamaa
	adherent	socialism
	Lakini kumpata tajiri mpenda Ujamaa ni shida sana. Kwa kweli utajiri na Ujamaa hupingana.
	‘But finding a rich man who loves socialism is quite unlikely. In truth wealth and socialism do not get along.’

44Leonard renders the whole passage as follows:

‘Poor men can have the souls of capitalists – parasites on their fellow men. Likewise, a rich man can value his wealth only because it can be used to help his fellow man… Likewise, a rich man can have a socialist spirit: He can value his wealth only because it can be used to help his fellow man… I have said that a rich man can be one who embraces socialism. But finding a rich man who loves socialism is quite unlikely. In truth wealth and socialism do not get along.’

45Nyerere is trying to show, with a certain amount of irony, that socialism is not favored by the rich. The poor, in some circumstances, may have a tendency to think in capitalist terms, that is, to act only for their own benefit, the rich never think outside their own interest. Leonard takes this passage to show that the propositions in the actual/narrative marked clauses are so marked because they are questioned, and so not taken to be true. However, we interpret Nyerere’s use of the actual/narrative here to be crucial to the rhetorical form of his argument, which is a reductio ad absurdum, i.e., one starts by accepting premises that one wants to refute, and then one shows that accepting those premises leads to false conclusions. We might argue that by accepting false premises as if they were true (using ka) we end up in an unreal world where rich people favor socialism. He then says that in fact there aren’t rich men who hold that view. If he used subjunctive e in these environments Nyerere would only be positing a world where ‘rich people favor socialism’ might be true, but that does not lead to the false conclusion if it turns out there are no such people, since the subjunctive e clausal marking only asserts that there might or might not be such people, which is why he concludes that wealth and socialism do not get along.

46Under our interpretation of this passage, the speaker is only committed to the truth of the ka propositions as a rhetorical tool where the subjunctive would not serve as well. We conclude that ka does not inherently signal the questioning of the event, as in Leonard’s theory (but see Section 4.2).

48However, Hopper considers the function of narrative morphemes in the so-called medial clauses to be a means of foregrounding a clause and backgrounding those clauses it is associated with, a line of analysis with a long pedigree (see also Longacre 1985; Dooley 2010; Schröder 2021, among others). Whether or not this approach is viable for other languages, for the Bantu languages in our limited sample (listed in Table 1), this approach does not explain the link between the narrative use and the interpretation of the ACT marking in complement clause contexts. Presumably, the claim would have to be that the complement is treated as foregrounded and the matrix clause is therefore backgrounded, but even if there were a clear test for backgrounded or foregrounded status, the causal link between the matrix clause event and the complement event would not be accounted for. Nothing in the foreground/background account predicts cause-and-effect interpretations nor does it directly lead to the possibility of consecutive interpretations. Moreover, it is not clear what it means for medial ka-clauses in a sequence to be backgrounded with respect to subsequent ka-clauses in the chain. While the Bantu markers discussed here may be a subcase of clause-chaining phenomena, the foregrounding/backgrounding approach is insufficiently specific to link the range of uses that actual clause morphology has in our Bantu sample.

4.2 Some unexplained cases

50Leonard (1980) points out a class of cases where the actual morpheme can mark matrix clause expressions of surprise or incredulity, as in the Kiswahili example he provides below.

(28)	Two thieves have robbed a man. In the course of the robbery, one kills a policeman. The accomplice says to the killer:
	Wewe	m‑jinga	sana.	Kwa	sababu	gani	u‑ka‑mw‑ua	yule
	you.sg	c1‑fool	real	for	c9.reason	what	sm.2sg‑narr‑om.c1‑kill	dem
	askari?
	c1.policemen
	‘You’re a fool! Why should you have killed that policeman?’ (Leonard 1980: 217, gloss added/adapted)

51This example, and other similar ones presented by Leonard, presume that the events the ka-clause describes have occurred, but have the flavor of “what the hell did you kill that policeman for?” (Leonard’s description). These are the cases that he describes as “questioning the event” because the inappropriateness of the event described is remarked upon, even though there is no question that the event has taken place. Our best guess for this usage within our approach is that the surprise consists in the absence of an appropriate contingent antecedent. The speaker treats the event described in the ka-clause as having taken place and the ka indicates there ought to have been a cause, a cause that the speaker cannot fathom.

52Another Kiswahili case pointed out by Leonard, one that is harder for us to talk our way around, concerns a dialog from a play wherein a woman has caught her husband seducing another woman. As part of an awkward dialog, she says (as rendered by Leonard):

(29)	i‑ngali‑kua	bora	u‑ka‑ni‑ach‑e
	expl‑cond‑be	better	sm.2sg‑narr‑om.2sg‑leave‑sbjv
	‘It would have been better if you left me.’

53Leonard remarks that the wife “is making what amounts to a prediction about what her husband will do and treats it as a polite command (to her husband).” However, one might also consider (29) as a necessary consequence of the unspoken situation, i.e., because he has cheated on her, he faces an inevitable (in her mind) consequence, namely, he must leave. On this suggested interpretation, ACT morphology is used because it is licensed by an implicit contingent antecedent, with the necessity of his leaving as a consequence. However, this seems a stretch and the occurrence of subjunctive morphology is, to us, unexpected.

54A Lubukusu example pointed out by Safir et al. (2020: 192) also remains unexplained (and was also unexplained in their paper). Such examples are also possible in Ekegusii.

(30)	a.	Wafula	a‑a‑nyw‑a	ka‑ma‑beele	ba‑ba‑ana	ba‑a‑lil‑a.
		Wafula	sm.c1‑act‑drink‑fv	c6‑c6‑milk	c2‑c2‑child	sm.c2‑act‑cry‑fv
		Wafula drank milk, so the children cried. [Lubukusu]
	b.	Wafula	a‑ka‑nyw‑a	a‑ma‑beere	a‑b‑ana	ba‑ka‑rer‑a.
		Wafula	sm.c1‑act‑drink‑fv	c6‑c6‑milk	c2‑c2‑child	sm.c2‑act‑cry‑fv
		Wafula drank milk, so the children cried. [Ekegusii]

55While one of the clauses might be the contingent antecedent for the other, that would still leave one clause without a contingent antecedent. We do not know why these are possible.

4.3 Summary of Section 4

57We summarize how the properties of the morphologically marked narrative use in Bantu listed in (31) align with our account.

58We have identified the morpheme described in (31a) as the actual morpheme. The properties in (31b, c) are consistent with our notion of dependency of the actual clause on a contingent antecedent. We predict (31d) based on the need for a contingent antecedent and the redundancy of marking an assertion with actual morphology, but as pointed out in Section 4.2, this leaves some apparent counterexamples unexplained. Property (31e) appears to be a consequence of the fact that actual-marked clauses do not behave as subjectless infinitives, and so nothing (such as control) requires that all the subjects in a chain should be identical.

59The link we are proposing between the complement and narrative uses of actual morphology is expected in our theory, as in both cases, the ka-clause is presented as true as a result of its contingent antecedent being true. In both contexts there can be a sequential reading (although sequence is not crucial, see Leonard 1980: 222). The cause-and-effect interpretation that we see in complement contexts thus bears close resemblance to the narrative use, where each clause in the chain describes an event or situation that temporally precedes the one that follows it.

60We acknowledge, however, that our account does not explain why actual clause propositions cannot denote an event or state of affairs that is not realized at the time of utterance. In discussions of narrative markers, it is usually just noted that they are typically used in narratives of past events (e.g. Nurse 2008), but this does not seem to be a sufficient explanation, especially in the actual clause complement context. Nonetheless, the fact that the narrative distribution of the actual morpheme and the complement distribution of the actual morpheme share the same semantic restriction supports our unified treatment.

5. Typology

62From what our research shows so far, the strongest statement, that a language can only synchronically have the actual complement effect if it has the narrative usage, seems to be supported. For every language in which we have found the actual clause complement effect, the same morphology is also used for clause-chaining. It is possible that some languages may have clause-chaining marked by a form that replaces tense, but without extending the use of that form to complement clauses, though we have no examples to offer at this time. More research will be necessary to determine what the best generalizations are.

63However, it could turn out that whatever the typological generalization is, it might have broader relevance not necessarily tied to overt tense-replacing morphology. Many clause-chaining languages have switch-reference markers. Switch reference morphology on a verb typically determines whether or not a given dependent clause in the chain has the same subject as an adjacent clause or has a different one. In languages where switch reference is the only morphology that indicates clause-chaining, it could be that switch reference in a complement clause (which has been taken to be typologically rare) could also signal an embedded assertion effect. Many questions remain open.

6. Conclusion

64We have argued that the morphology we call “actual” and its counterparts in other Bantu languages where it is called “narrative” or “consecutive” signals that UTT is committed to the actual-clause being true as a consequence of its contingent antecedent being true. We have explored the relationship between the actuality reading in actual clause complements and the use of actual morphology in narrative on this basis. The claim that actual clauses are assertions relates both of its uses to the moment of speech, which entails that UTT presents as true a proposition that is not presupposed to be true. By syntactic tests, we showed that actual clauses can indeed be complements, even though they function as assertions. The complement clause usage is thus predicted to be incompatible with factive contexts, which we showed to be the case. We speculate that the complement clause usage is adapted from the narrative usage, which would account for the same morphology appearing in both contexts. Insofar as not every language with dedicated narrative/actual morphology also allows clauses of that type to be complement clauses, the implication is only one way, i.e., if ‘actual’ is a complement clause type, then the language uses actual morphology for narrative clause-chaining (but not vice versa). We then expect that there will be languages with a special clause type for clause-chaining that may not use the same morphology for complement clauses. However, our empirical research is ongoing.

Acknowledgments

65We acknowledge NSF‑BCS‑1829122 which funds the Afranaph Project (Ongoing) that has provided major support for our research. We are grateful for useful feedback derived from presentations at the Afranaph Development Workshop III at Georgetown University and the workshop on Bantu universals at WOCAL 10. Reviewers of an earlier draft of this paper were especially helpful and their comments led to significant revisions. We also thank our consultants Deo Kawalya and Saudah Namyalo (Luganda), Harriet Kasande (Rutooro), Samuel Obuchi and Roseline Kerubo Speke (Ekegusii), John Bosco Gatabazi (Kirundi), Amani Lusekelo and unnamed University of Dar es Salaam lecturers (Kiswahili), Mohamed Mraja and Suleiman Chembea (Kidigo), Bhoke Tabitha (Kikuria), and Christine Katunge Muthini (Kikamba). The first and second authors provided the Lubukusu data. The second author also provided Kiswahili data.

Abbreviations

act	actual marker
am	associative marker
appl	applicative marker
c	noun class, typically followed by a noun class number, e.g. c6‑
caus	causative
comp	complementizer
cond	conditional
dem	demonstrative
dep	dependent
disj	disjoint marker
expl	expletive
foc	focus
fut	future
fv	final vowel
inf	infinitive
narr	narrative
neg	negation
om	object marker
pass	passive
pfv	perfective
pst	past
sbjv	subjunctive
sg	singular
sm	subject marker
subj	subjunctive
tns	tense

References

66Anand, Pranav & Valentine Hacquard. 2014. Factivity, belief and discourse. In Luka Crnič & Uli Sauerland (eds.), The art and craft of semantics: A festschrift for Irene Heim, vol. 1, 69‑90. Cambridge, MA: MIT Working Papers in Linguistics.

67Dahl, Östen. 1977. Games and models. In Östen Dahl (ed.), Logic, pragmatics and grammar, 147‑200. Gothenburg: Department of Linguistics, University of Gothenburg.

68Dahl, Östen. 1985. Tense and aspect systems. Oxford: Blackwell.

69Doke, C. M. & S. M. Mofokeng. 1957. Textbook of Southern Sotho grammar. Cape Town: Maskew Miller Longman.

70Dooley, Robert A. 2010. Exploring clause chaining. SIL Electronic Working Papers 2010‑001. 1‑20 (accessed 24 May 2023).

71Englebretson, Robert & Wambũi Mũringo Wa-Ngatho (eds.). 2015. A basic sketch grammar of Gĩkũyũ. Rice Working Papers in Linguistics 6 (Special issue) (accessed 11 May 2023).

72Haspelmath, Martin. 1995. The converb as a cross-linguistically valid category. In Martin Haspelmath & Ekkehard König (eds.), Converbs in cross-linguistic perspective: Structure and meaning of adverbial verb forms – adverbial participles, gerunds, 1‑55. Berlin: Mouton de Gruyter.

73Hopper, Paul. 1979. Aspect and foregrounding in discourse. In Talmy Givón (ed.), Syntax and Semantics 12, 213‑241. New York, NY: Academic Press.

74Johnson, Bo. 1979. Hebräisches Perfekt und Imperfekt mit vorangehendem w^e. Lund: Gleerup.

75Keach, Camillia. 1995. Subject and object markers as agreement and pronoun incorporation in Swahili. In Akinbiyi Akinlabi (ed.), Theoretical approaches to African Linguistics, 109‑116. Trenton, NJ: African World Press.

76Khoali, B. Thakampholo. 1991. A Sesotho tonal grammar. Champaign, IL: University of Illinois at Urbana-Champaign (PhD dissertation).

77Kiparsky, Paul & Carol Kiparsky. 1970. Fact. In Manfred Bierwisch & Karl Erich Heidolph (eds.), Progress in linguistics, 143‑173. Berlin: De Gruyter Mouton.

78Leonard, Robert A. 1980. Swahili e, ka, and nge as signals of meanings. Studies in African Linguistics 11(2). 209‑226.

79Letšeng, ‘Makhauta C. 1995. La structure tonale du verbe en sotho du sud. Grenoble: Université Stendhal (M.A. thesis).

80Longacre, Robert E. 1985. Sentences as combinations of clauses. In Timothy Shopen (ed.), Language typology and syntactic description. Vol. 2: Complex constructions, 235‑286. Cambridge: Cambridge University Press.

81Matić, Dejan & Daniel Wedgwood. 2013. The meanings of focus: The significance of an interpretation-based category in cross-linguistic analysis. Journal of Linguistics 49(1). 127‑163.

82Nurse, Derek. 2008. Tense and aspect in Bantu. Oxford: Oxford University Press.

83Posthumus, Lionel C. 1991. Past subjunctive or consecutive mood? South African Journal of African Languages 11(3). 91‑95.

84Riedel, Kristina, Hannah Sarvasy & Katherine Demuth. 2019. The narrative past inflection in Sesotho child and child-directed speech. Frontiers in Communication 69(4). 1‑17.

85Safir, Ken, Justine Sikuku & Mark Baker. 2020. Actual clauses in Lubukusu. In Eno-Abasi Urua, Francis Egbokhare, Oluseye Adesola & Harrison Adeniyi (eds.), African languages in time and space: A festschrift in honour of Professor Akinbiyi Akinlabi, 189‑206. Ibadan: Zenith BookHouse.

86Schröder, Helga. 2013. S/A pivot in Toposa clause chaining. Paper presented at the Nilo-Saharan Linguistics Colloquium, Cologne, 22‑24 May 2013, ms.

87Schröder, Helga. 2021. The syntax and semantics of clause-chaining in Toposa. Studies in African Linguistics 49(1). 60‑83.

88Simons, Mandy. 2007. Observations on embedding verbs, evidentiality, and presupposition. Lingua 117(6). 1034‑1056.