1. Overview

2This special issue is made up of five contributions from the workshop (Dom, Bar-el, Kanijo & Petzell; Laine, Bernander & Aunio; Meyer; Treis; Yoneda) and three additional papers (Creissels; Schneider-Blum; Allassonnière-Tang, Robert & Voisin). These eight studies cover a total of 73 African languages from all over the continent. Table 1 lists these languages using glossonyms adopted in the respective papers, as well as their ISO 639-3 codes, language families as classified by Glottolog (Hammarström et al. 2022), and the authors of the papers in which each language is covered.

Table 1 — Overview of the languages covered in the contributions

Glossonym	Code	Language family	Authors
Adamawa Fula	fub	Atlantic-Congo	Creissels
Afar	aar	Afro-Asiatic	Creissels
Amharic	amh	Afro-Asiatic	Creissels
Balant Ganja	bjt	Atlantic-Congo	Creissels; Allassonnière-Tang et al.
Balant Kəntohe	ble	Atlantic-Congo	Allassonnière-Tang et al.
Bambara	bam	Mande	Creissels; Allassonnière-Tang et al.
Basari	bsc	Atlantic-Congo	Allassonnière-Tang et al.
Baule	bci	Atlantic-Congo	Creissels
Bedik	tnr	Atlantic-Congo	Allassonnière-Tang et al.
Bobo	bbo	Mande	Allassonnière-Tang et al.
Emai	ema	Atlantic-Congo	Creissels
Fon	fon	Atlantic-Congo	Creissels
Ful of Massina	ffm	Atlantic-Congo	Allassonnière-Tang et al.
Gbaya	gya	Atlantic-Congo	Creissels
Hausa	hau	Afro-Asiatic	Creissels
Herero	her	Atlantic-Congo	Creissels; Yoneda
Humburi Senni	hmb	Songhay	Creissels
Ikoma	ntk	Atlantic-Congo	Dom et al.; Laine et al.
Ishenyi	ntk	Atlantic-Congo	Laine et al.
Jamsay	djm	Dogon	Creissels
Jóola Banjal	bqj	Atlantic-Congo	Allassonnière-Tang et al.
Jóola Fóoñi	dyo	Atlantic-Congo	Creissels; Allassonnière-Tang et al.
Jóola Karon	krx	Atlantic-Congo	Allassonnière-Tang et al.
Jóola Kasa	csk	Atlantic-Congo	Allassonnière-Tang et al.
Jóola Keeraak	hhr	Atlantic-Congo	Allassonnière-Tang et al.
Jóola Kwaatay	cwt	Atlantic-Congo	Allassonnière-Tang et al.
Kagbaaga	bjg	Atlantic-Congo	Allassonnière-Tang et al.
Kagulu	kki	Atlantic-Congo	Dom et al.
Kakabe	kke	Mande	Creissels; Allassonnière-Tang et al.
Kambaata	ktb	Afro-Asiatic	Treis
Kanuri	knc	Saharan	Creissels
Kisi	kqs	Atlantic-Congo	Allassonnière-Tang et al.
Konyagi	cou	Atlantic-Congo	Allassonnière-Tang et al.
Koroboro Senni	ses	Songhay	Creissels
Kupsabiny	kpz	Nilotic	Creissels
Kutu	kdc	Atlantic-Congo	Dom et al.
Kwere	cwe	Atlantic-Congo	Dom et al.
Laalaa	cae	Atlantic-Congo	Allassonnière-Tang et al.
Landuma	ldm	Atlantic-Congo	Allassonnière-Tang et al.
Lingala	lin	Atlantic-Congo	Creissels
Luguru	ruf	Atlantic-Congo	Dom et al.
Mandinka	mnk	Mande	Creissels; Allassonnière-Tang et al.
Maninka of Niokolo	mku	Mande	Allassonnière-Tang et al.
Manjaku of Bassarel	mfv	Atlantic-Congo	Allassonnière-Tang et al.
Mano	mev	Mande	Creissels
Matengo	mgv	Atlantic-Congo	Yoneda; Dom et al.
Minyanka	myk	Atlantic-Congo	Creissels
Muher	sgw	Afro-Asiatic	Meyer
Nalu	naj	Atlantic-Congo	Allassonnière-Tang et al.
Nata	ntk	Atlantic-Congo	Laine et al.
Ngoreme	ngq	Atlantic-Congo	Laine et al.
Nyun Gubëeher	bab	Atlantic-Congo	Allassonnière-Tang et al.
Nyun Gujaher	ccj	Atlantic-Congo	Allassonnière-Tang et al.
Nyun Guñaamolo	bcz	Atlantic-Congo	Allassonnière-Tang et al.
Pajaade	bif	Atlantic-Congo	Allassonnière-Tang et al.
Palor	fap	Atlantic-Congo	Allassonnière-Tang et al.
Pepel	pbo	Atlantic-Congo	Allassonnière-Tang et al.
Pulaar (Ful of Futa Toro)	fuc	Atlantic-Congo	Allassonnière-Tang et al.
Saafi	sav	Atlantic-Congo	Allassonnière-Tang et al.
Sandawe	sad	Sandawe	Creissels
Sar	mwm	Central Sudanic	Creissels
Seereer	srr	Atlantic-Congo	Creissels; Allassonnière-Tang et al.
Sherbro	bun	Atlantic-Congo	Allassonnière-Tang et al.
Sidaama	sid	Afro-Asiatic	Creissels
Soninke (of Kingi)	snk	Mande	Creissels; Allassonnière-Tang et al.
Soninke of Bakel	skq	Mande	Allassonnière-Tang et al.
Soso	sus	Mande	Allassonnière-Tang et al.
Swahili	swa	Atlantic-Congo	Creissels; Yoneda; Dom et al.
Temne	tem	Atlantic-Congo	Allassonnière-Tang et al.
Tima	tms	Katla-Tima	Schneider-Blum
Tswana	tsn	Atlantic-Congo	Creissels
Wolof	wol	Atlantic-Congo	Creissels; Allassonnière-Tang et al.
Zalamo	zaj	Atlantic-Congo	Dom et al.

3In the remainder of this Introduction, we provide a definition of the noncausal/causal alternation (Section 2), present a state-of-the-art of the typological research on this topic (Section 3), and compare the dominant typologies of correspondence types (Section 4) and data collection methodologies (Section 5). Section 6 closes this Introduction with an overview of the contributions.

2. Defining the noncausal/causal alternation

(1)	The ship sank.
	a.	role frame:	Y	sank
			|
	b.	coding frame:	Arg1-nom	V.sbj[1]

(2)	The torpedo sank the ship.
	a.	role frame:	X	sank	Y
			|		|
	b.	coding frame:	Arg1-nom	V.sbj[1]	Arg2-acc

5The role frame indicates the semantic participants, and the coding frame presents the syntactic arguments. The noncausal event sink (intr.) entails one semantic role, represented as Y in the role frame in (1a), which is the participant undergoing the sinking event. Syntactically, the semantic role Y is coded by the subject argument of the noncausal verb sink (intr.), represented as Arg1-nom in the coding frame in (1b). The mapping of the semantic role Y onto the subject argument (Arg1-nom) is represented by the vertical line connecting the role and coding frames. The causal event sink (tr.) has two semantic roles, X and Y in (2a). The semantic role Y in (2a) corresponds to the semantic role Y in (1a). In the argument structure of the causal predicate sink (tr.), though, the semantic role Y is coded by the object argument (2b), represented as Arg2-acc, and the subject argument (Arg1-nom) encodes the semantic role X, which is the causer participant.

6The valency of corresponding noncausal and causal predicates differ from each other in two important ways: (i) compared to the noncausal predicate, the causal predicate has one more argument in its argument structure, namely, the causer, and (ii) the affected participant of the event encoded by the subject argument of the noncausal predicate is encoded by the object argument of the causal predicate. While noncausal predicates are generally monovalent and the corresponding causal predicates are bivalent, there are bivalent noncausal predicates with trivalent causal counterparts, such as English learn/teach in (3).

(3)	a.	The children learned a song.
	b.	The clown taught the children a song.

7The semantics of the noncausal/causal alternation and the causative voice are so closely related that a clear distinction in semantic terms is difficult to establish (we refer the reader to Haspelmath et al. 2014; Haspelmath 2016; Zúñiga & Kittilä 2019: 15‑20). The crucial feature distinguishing the noncausal/causal alternation from the causative voice is the cross-linguistic variation in the coding of valency change. The coding strategies of valency change on predicates participating in the noncausal/causal alternation exhibit major cross-linguistic variability, whereas the causative voice is coded only through valency-increasing devices (Haspelmath 1993: 92; Haspelmath et al. 2014: 590‑591). Some event and predicate types have a strong tendency for causative derivation cross-linguistically, e.g. wash (tr.)/make someone wash or cut (tr.)/make someone cut, while other event and predicate types exhibit major variation in the formal correspondences between the noncausal and causal events, e.g. wake up (intr./tr.) or hang (intr./tr.) (for detailed studies on the underlying determinants of coding variation, see Haspelmath et al. 2014; Haspelmath 2016). The schemas in (1)‑(2) above illustrate two correspondences for the noncausal/causal alternation in English: (i) noncausal and causal events are expressed by the same predicate, e.g. sink (intr./tr.) and (ii) noncausal and causal events are expressed by lexically unrelated predicates, e.g. learn/teach.

8Comparative studies such as Nedjalkov (1969), Haspelmath (1993) and Nichols et al. (2004) typologize the coding variation attested in the world’s languages. This type of formal variation in correspondences is also the focus of the contributions in this special issue. We present a brief overview of these typological studies in the following section.

3. Typological studies on the noncausal/causal alternation

9Typological research on transitivity alternations like the noncausal/causal alternation goes back to Nedjalkov (1969), who examines the formal derivation between the intransitive and transitive members of verb pairs denoting the following four meanings in 60 languages.

(4)	a.	laugh
	b.	make laugh; amuse

(5)	a.	(come to a) boil; be(come) boiling
	b.	boil something

(6)	a.	become burned; burn (intr.)
	b.	burn something

(7)	a.	become broken; break (intr.)
	b.	break something

10The difference between the members of the verb pairs in (4)‑(7) relates mainly to a difference in transitivity (intransitive vs. transitive) and semantic causativity (noncausal vs. causal). In each of the four verb pairs, the verb in (a) is noncausal and intransitive, while the verb in (b) is causal and transitive.

11Building on the studies of Nedjalkov (1969) and Nedjalkov & Silnitsky (1973), Haspelmath (1993) investigates the formal relationships between the members of so-called “inchoative/causative” verb pairs. Inchoative verbs are defined by Haspelmath (1993) as verbs denoting change-of-state or change-of-location situations that occur spontaneously, e.g. sink (intr.). Two decades later, Haspelmath et al. (2014) and Haspelmath (2016) investigate a much broader type of verb alternation, which they call the “causal-noncausal alternation”. This new classification includes any verb pair in which the members are related by means of a semantic causative relation. The noncausal member can be any verb to which a causal meaning can be added, such as sink (intr.) > sink something, laugh > make someone laugh, or cut something > make someone cut something. Note that the latter two verb pairs are not “inchoative/causative” according to the semantic definition from Haspelmath (1993).

12Nichols et al. (2004) and Grünthal & Nichols (2016) constitute another line of research that draws on the seminal work by Nedjalkov (1969). Nichols et al. (2004) refer to the noncausal member as the “plain” verb and to the causal member as “induced”. In contrast, Grünthal & Nichols (2016) use the terms “non-causative” and “causative” for members of the “causative alternation”. They characterize this verb alternation in terms of the following features:

“semantics (‘scare, frighten’ adds causation by an agent or force), argument structure (the agent or force is the A of the causative verb, and the S or A of the non-causative becomes the causee of the causative verb), valence or argument coding (the causative verb is usually transitive: the new A is usually its subject and the causee is usually an object), and often the morphology of the verb (the causative verb may have a causative affix or other marker of derived transitivity, and/or the non-causative one may have a marker of derived intransitivity)” (Grünthal & Nichols 2016: 12; bolding added).

13The type of verb alternation investigated by Nichols et al. (2004) and Grünthal & Nichols (2016) is less restricted than the inchoative/causative alternation of Haspelmath (1993) but not as broad as the causal-noncausal alternation of Haspelmath et al. (2014) and Haspelmath (2016).

14In summary, while these studies seem to focus on the same type of verb alternation, they in fact deal with different verb types and/or transitivity alternations. For this special issue, we adhere to the definition of the noncausal/causal alternation provided in Section 2. In the following section, we present a brief overview of two typologies of correspondences for noncausal/causal pairs.

4. Correspondence types

15All contributions in this issue make use of either Haspelmath’s (1993) or Nichols et al.’s (2004) typology of correspondence types to identify and categorize the formal relations observed in the respective African languages. In this section, we give a brief overview of these two typologies.

16Based on a set of 31 noncausal/causal predicates in a sample of 21 languages, Haspelmath (1993: 91‑92) provides a typology of five correspondence types: causative, anticausative, equipollent, suppletive and labile. In (8)-(12), we provide definitions and examples of each type (examples are taken from Haspelmath (1993: 90-92).

(8)	Causative
	The causal predicate is formally coded by an affix, auxiliary or stem modification that is not present on the noncausal predicate.
	Georgian:	duɣ-s	‘cook (intr.)’
		a-duɣ-ebs	‘cook (tr.)’

(9)	Anticausative
	The noncausal predicate is formally coded by an affix, auxiliary or stem modification that is not present on the causal predicate.
	Russian:	katat’-sja	‘roll (intr.)’
		katat’	‘roll (tr.)’

(10)	Equipollent
	Both the noncausal and causal predicates are derived from the same stem, formally coded by different affixes, different auxiliaries, or different stem modifications.
	Japanese:	atum-aru	‘gather (intr.)’
		atum-eru	‘gather (tr.)’

(11)	Suppletive
	The noncausal and causal predicates are lexically unrelated.
	Russian:	goret’	‘burn (intr.)’
		žeč’	‘burn (tr.)’

(12)	Labile
	A single predicate denotes both the noncausal and causal meanings.
	Modern Greek:	svíno	(i) ‘go out’
			(ii) ‘extinguish’

17The term ‘causative’ in (8) is more commonly used to refer to the causative voice. However, with respect to the noncausal/causal alternation, it refers to the correspondence type in which the causal verb is derived through a transitivization process that is absent on the noncausal verb. Although the formal coding of a noncausal/causal pair in causative correspondence is often the same as the formal coding of the causative voice in a given language, this is not always the case. For example, in Lai the causal verb tsoon-piak ‘teach’ is derived from noncausal tsoon ‘learn’ and thus in causative correspondence. However, the causal predicate is formally coded for the benefactive applicative voice and not causative voice (Nichols et al. 2004: 152). Likewise, the term anticausative is more commonly used to refer to a voice alternation (Moreno 1985; Haspelmath 1987; Alexiadou et al. 2006). However, the anticausative correspondence type describes other overtly coded valency-decreasing voices, such as middle, passive and/or reflexive.

18Nichols et al. (2004: 159) propose a typology of nine correspondence types, based on a set of 18 noncausal/causal predicates in a sample of 80 languages. In addition to referring to some of Haspelmath’s correspondence types by different labels (e.g. Haspelmath’s causative, anticausative and labile are labeled by Nichols et al. 2004 as augmented, reduced and ambitransitive, respectively), Nichols et al. (2004) distinguish four correspondence types that are subsumed under the equipollent correspondence type in Haspelmath (1993). These four are defined in (13)-(16), illustrated with examples from Nichols et al. (2004: 159-160).

(13)	Double derivation
	Both the noncausal and causal predicates are formally derived by a detransitivization and transitivization process, respectively.
	Siberian Yupik:	aghagh-nga-	‘hang (intr.)’
		aghagh-te	‘hang (tr.)’

(15)	Auxiliary change
	The same predicate is used with different auxiliaries to express the noncausal and causal meanings.
	Ingush:	ieghaz-v.uoda (angry-go)	‘become angry’
		ieghaz-v.ug (angry-lead)	‘anger, make angry’

(16)	Ablaut
	The noncausal and causal predicates are distinguished by a difference in vowel or consonant grade in the stem.
	Lai:	Ɂa-thin phaaŋ	‘be afraid’
		Ɂa-thin phaɁn	‘frighten’

19Creissels (this issue) shows examples from multiple African languages of ablaut pairs where the difference between the two members is encoded by a difference in tone instead of vowel or consonant grade. This is illustrated in (17).

(17)	a.	Pere:	nál	‘melt (intr.)’
			nàl	‘melt (tr.)’
	b.	Hausa:	yàwaità (v3)	‘increase (intr.)’
			yawàitā (v1)	‘increase (tr.)’
	c.	Humburi Senni:	céyrí	‘break (intr.)’
			cèyrì	‘break (tr.)’

20Finally, Nichols et al. (2004) introduce an adjective correspondence type that is not included in Haspelmath (1993).

(18)	Adjective
	The noncausal meaning is denoted by an adjectival predicate and the causal meaning by a verbal predicate.
	English:	be afraid
		frighten

21Nichols et al. (2004: 153) propose that more than one correspondence can hold between the members of a noncausal/causal pair. For example, in Zaramo the verbs mal/mambuk-iz ‘finish (intr.)/finish (tr.)’ are in a suppletive correspondence, i.e. the verbs have different roots. In addition, the pair is also in causative correspondence in that the causal verb mambuk-iz ‘finish (tr.)’ is derived with the transitivizing causative suffix -iz. Similarly, the same root or stem can participate in two (or more) different correspondence types. In Bambara, the root sɛ̀gɛn is used to denote the noncausal/causal meanings ‘get tired/make tired’. The noncausal/causal meanings can be denoted by using the root in a labile correspondence, as in (19a), or in a causative correspondence, as in (19b).

(19)	Bambara
	a.	labile:	sɛ̀gɛn	‘get tired’
			sɛ̀gɛn	‘make tired’
	b.	causative:	sɛ̀gɛn	‘get tired’
			lá-sɛ̀gɛn	‘make tired’

22Haspelmath and/or Nichols et al.’s typologies of correspondence types are examined in each of the contributions in this special issue. In the following section we summarize the methodologies used for these cross-linguistic studies.

5. Methodology and data collection

23For their typological studies of the noncausal/causal alternation, Haspelmath (1993) uses a list of 31 meanings while Nichols et al. (2004) compare a set of 18 meanings across the languages in their sample. The two lists are given in Table 2. They share 8 meanings, which are listed first in Table 2. All papers in this special issue examine datasets based on one or both of these lists. The studies by Creissels, Treis and Schneider-Blum also include additional noncausal/causal meanings.

Table 2 — An overview of the lists of noncausal/causal meanings used by Haspelmath (1993) and Nichols et al. (2004)

	Haspelmath (1993)	Nichols et al. (2004)
1	die/kill	die/kill
2	learn/teach	learn, know/teach
3	boil (intr./tr.)	(come to) boil/(bring to) boil
4	burn (intr./tr.)	burn, catch fire/burn, set fire
5	break (intr./tr.)	break (intr./tr.)
6	open (intr./tr.)	open (intr./tr.)
7	dry (intr./tr.)	dry/make dry
8	turn (intr./tr.)	turn over (intr./tr.)
9	wake up (intr./tr.)	laugh/make laugh, amuse, strike as funny
10	close (intr./tr.)	sit/seat, have sit, make sit
11	begin (intr./tr.)	eat/feed, give food
12	gather (intr./tr.)	see/show
13	spread (intr./tr.)	be(come) angry/anger, make angry
14	sink (intr./tr.)	fear, be afraid/frighten, scare
15	change (intr./tr.)	hide, go into hiding/hide, conceal, put into hiding
16	melt (intr./tr.)	be(come) straight/straighten, make straight
17	be destroyed/ destroy (intr./tr.)	hang/hang (up)
18	get lost/lose	fall/drop, let fall
19	develop (intr./tr.)
20	connect (intr./tr.)
21	rock (intr./tr.)
22	go out/put out
23	rise/raise
24	finish (intr./tr.)
25	roll (intr./tr.)
26	freeze (intr./tr.)
27	dissolve (intr./tr.)
28	fill (intr./tr.)
29	improve (intr./tr.)
30	split (intr./tr.)
31	stop (intr./tr.)

24Haspelmath’s list of 31 meanings has been adopted by The World Atlas of Transitivity Pairs (2014), an online database that expands Haspelmath’s language sample. Despite the value of these meaning lists for cross-linguistic exploration, there are nevertheless several methodological issues that arise. The studies presented in this special issue address some of these methodological problems, which we summarize here.

25First, complications of synonymy arise when multiple verbs are attested for a single meaning, as shown for Kambaata in (20). Note that these synonyms do not all have the same correspondence.

(20)	Kambaata synonymous noncausal/causal pairs (Treis, this issue)
		Verb pair	Meaning
	a.	bínn=y-/bínn=a’-	‘spread, scatter (solids) (intr./tr.)’
	b.	finc-am-/finc-	‘spread, scatter (solids) (intr./tr.)’
	c.	iill-/iill-is	‘spread (e.g. news) (intr./tr.)’
	d.	tar-/tar-siis-	‘spread (e.g. disease) (intr./tr.)’

26The opposite also occurs, where one predicate or predicate pair denotes more than one of the meanings in the list. This is illustrated for Muher in (21). The noncausal verb tänässam means both ‘wake up (intr.)’ and ‘rise’, alternating with two different causal predicates.

(21)	Muher homonymous predicate occurring in two noncausal/causal pairs (Meyer, this issue)
	a.	tänässam/k’ɨsäkk’äsäm	‘wake up (intr.); rise/wake up (tr.)’
	b.	tänässam/nässam	‘wake up (intr.); rise/raise’

6. A summary of the contributions in this special issue

28The following is a summary of the eight contributions in this issue, in order of their appearance.

29Creissels investigates the frequency of five correspondence types in a sample of 13 noncausal/causal pairs in 30 sub-Saharan languages. The sample is restricted to pairs whose noncausal member is monovalent, denoting a process (not a state) typically undergone by a concrete inanimate entity and easily construed as occurring without the involvement of an external causer. The main finding that emerges from this study is that 10 out of the 30 African languages have a predominance for ambitransitivity (lability), which is unmatched by any of the other four alternation types. This observation raises two questions: 1) Is ambitransitivity used for other verb alternations in these 10 languages, and 2) what explains the shared pattern between these 10 languages (and, possibly, other languages beyond this study)? Creissels explores the first question by examining the occurrence of ambitransitivity for three additional verb alternation types in the 10 languages. Remarkably, only three languages exhibit an extended prominence of ambitransitivity when taking into account the three additional verb alternation types. The explanation of the shared predominance of ambitransitivity in these languages requires further research.

30Treis explores the variation in coding strategies for the noncausal/causal alternation in Kambaata, a Cushitic language of Ethiopia. Based on Haspelmath’s list of 31 verbal concepts, this study reveals that the predominant correspondence type in Kambaata is causative, with anticausative a close second. However, Treis shows that if we take into account all verbal and ideophonic lexemes used in Kambaata to express noncausal/causal alternations, the significance of the equipollent strategy is revealed in the language. None of the verbs in equipollent correspondence are captured if the analysis is restricted to Haspelmath’s list. The Kambaata data also highlights that lability is an uncommon strategy in this language, despite being analyzed as the most common correspondence type in Creissels’ (this issue) study. Even when taking into account the additional lexemes, Kambaata still exhibits only a single labile verb. Treis’ contribution sets the stage for future comparisons of noncausal/causal alternations in other Cushitic languages.

31Meyer analyses the correspondence types of noncausal/causal pairs in Muher, a Semitic language spoken in Ethiopia. Using Haspelmath’s (1993) typology of five correspondence types and list of 31 noncausal/causal meanings, the study reveals that causativization is the most frequently attested correspondence type in Muher, followed by anticausativization, then suppletion, equipollence and, finally, lability. Muher has a set of three productive derivational prefixes coding valency change, mediopassive t(ä)‑, direct causative a‑, and indirect causative at‑, which motivates the prevalence of causativization and anticausativization in noncausal/causal verb pairs. Muher also has a very productive morphosyntactic construction, the impersonal, which can be employed to background an agent or causer and put the patient in focus. The argument structure of an impersonal construction is transitive, with a generic third person subject and the patient coded as object. Therefore, impersonal verbs cannot be analysed as noncausal verbs, whose role and coding frames only include one argument denoting the patient role. Thus, pairs of impersonal/transitive verbs are not analysed as coding the noncausal/causal alternation.

32Allassonnière-Tang et al.’s contribution provides a statistical analysis of the distribution of coding strategies of valence orientation in 38 languages belonging to the Mande, Atlantic and Mel groups, and spoken in and around Senegal. Allassonnière-Tang et al. use machine-learning methods (clustering and tree-based computational classifiers) to investigate the connection between valence orientation and genealogical affiliation. Their analyses show that each linguistic family has rather stable trends of valence orientation, which in turn can predict the genealogical affiliation of a language with an accuracy above the majority baseline.

33Schneider-Blum’s contribution discusses the noncausal/causal alternation in Tima, a language from the Nuba mountains in Sudan. The study has two goals. First, it examines a correlation of the verbal derivation marking options in Tima with different noncausal/causal alternations described by Nichols et al. (2004). The study establishes that Tima verbal derivations can be classified into three of Nichols et al.’s correspondence types: augmentation, double derivation and reduction. These correspondence types are further subdivided into different groups based on the semantic roles of the subject and object and verb-inherent semantics. The second goal is to scrutinize the grammaticalization path of the causative marker in Tima, which is used for overlapping functions. Schneider-Blum shows that in Tima, the grammaticalization path proceeds from causative via reflexive/anticausative to passive. This deviates from a grammaticalization path suggested in Haspelmath (1990), which indicates a change from causative via reflexive-causative to passive. In Tima, a reflexive-causative function is not attested.

34Yoneda’s study examines the noncausal/causal verb alternation in Swahili, a Bantu language spoken widely in East Africa. The study is based on Haspelmath’s (1993) 31 verb pairs. In contrast to Haspelmath, who classifies Swahili as ambiguous between a causativizing and an anticausativizing language, Yoneda’s study firmly classifies Swahili as a causativizing language. One reason for the difference between the two classifications lies in the treatment of data that involves the suffixes -(i)k for noncausal verbs and -(i)sh or -(i)z for causal verbs, as in chemka ‘boil (intr.)’ vs. chemsha ‘boil (tr.)’. In Haspelmath’s analysis, verb pairs that involve these changes are in equipollent correspondence, in which the noncausal and causal verbs are derived from the same root. In Yoneda’s study, this data is analysed as evidence for causative correspondence. This is due to a diachronic phonological process involving the suffixation of -y to the stem, resulting in a fricative of the preceding consonant, e.g., k + y > sh in chemka ‘boil (intr.)’/chemk-y-a > chemsha ‘boil (tr.)’. This sound change is known as spirantization in Bantu linguistics. Another distinction between Haspelmath’s and Yoneda’s classifications is that Haspelmath classifies some verbs as belonging to the same lexeme. However, these verbs are separate lexical items, each having a different meaning.

35Dom et al. focus on the same sound change as Yoneda, i.e. spirantization, in a group of eight East Bantu languages: Ikoma, Kagulu, Kutu, Kwere, Luguru, Matengo, Swahili and Zaramo. In particular, they examine several verb pairs whose causal members are reflexes of verbs derived with a causative suffix reconstructed to Proto-Bantu as *-i. These verb pairs used to be in a causative correspondence. However, the authors propose that due to morphophonological changes involving reflexes of the Proto-Bantu causative suffix *-i, the formal relationships between these verb pairs changed from causative to other correspondence types. The analysis presented by Dom et al. contrasts with Yoneda (this issue), with two alternative approaches to noncausal/causal verb pairs exhibiting spirantization in East Bantu.

36Laine et al.’s contribution examines the noncausal/causal alternation in four Bantu languages, Ikoma, Ishenyi, Nata and Ngoreme, which form a closely knit genetic unit called Western Serengeti (WS). The study shows the strong preference for the causative strategy in these languages. Laine et al. compare this preference with cases in other Bantu languages of Tanzania but also examine the characteristics of WS in a wider typological context. The data shows that WS often makes use of the reflexive prefix in deriving anticausative verbs instead of the neuter derivational suffix more commonly used in other East Bantu languages, but which is unproductive in WS.

Abbreviations

acc	accusative case
Arg	syntactic argument
intr.	intransitive
nom	nominative case
sbj	subject
tr.	transitive
V	verb