2The paper is organized as follows. Section 1 gives a sketch of the sociolinguistic situation. Section 2 provides some preliminary observations of the interdialectal differences in phonology. Section 3 gives some details on morphosyntactic variation. Section 4 presents differences in lexicon. The results are discussed in Section 5 where I explain, in particular, why some of the dialectal differences could be in fact an issue of contact with different languages: Kpelle, especially in the North of the Mano zone, and Dan in the South.

Sociolinguistic situation

4In Guinea, Mano counts three “dialects”: Zaan (zà̰à̰), the easternmost “dialect” spoken around the Bossou town, Maa (màá), the central “dialect” spoken in the city of Nzérékoré and to the south of it, and Kpeinson (kpéŋ́sɔ̰̀), the South‑Western dialect spoken around the town of Diécké. The three varieties are fairly close to one another, Kpeinson being the closest to the Liberian variety spoken around Ganta. Diécké and Ganta are just a few kilometers apart, and there is quite intense communication between the two towns.

5In Liberia, the dialectal variation is much more salient. According to the account by Zetterström, “[t]here are several different dialects of the Mano language spoken within Liberia. One dialect boundary goes through the northern part of the [Yamein] clan. Those who live north of this boundary speak like the Guinea‑Mano and this dialect is obviously strongly influenced by the Mandingo‑language. South of Sanniquelli there is another boundary, south of which there are at least two different dialects. The differences between these dialects are considerable and the Mano of the south often have difficulties in understanding those of the north” (Zetterström 1976: 16‑17). The dialectal classification provided by Liberian Mano themselves partially confirms Zetterström’s observations: the Northern dialect, Maalaa (máá lāā), spoken around Sanniquellie; the Central dialect, Maazein (máá zèŋ́) spoken in Ganta, as well as in Kpein and Flumpa, two of the three villages that I visited; and the Southern dialect, Maabei (máá bèí), spoken in Saklepea and in Gbanquoi, the third village I visited. The Northern and the Central dialects are fairly similar. Mano literacy materials and, very likely, the Bible translation are based on these two dialects. The Southern dialect is the most distinctive of all three and as compared to Guinean varieties. While the Northern and the Central varieties occupy about a third of the Mano territory, it is unlikely that the Southern variety will cover the remaining two thirds. Therefore, it is expected that there might be several other varieties spoken to the south of Gbanquoi.

6While doing a dialectal survey of Liberian varieties I traveled with Pe Mamy, my primary language consultant, a speaker of the central Guinean dialect, Maa. Our guide was Leelamen Zarwolo, a speaker of the Central Liberian dialect and native of Flumpa, who also works to promote literacy in Mano and is affiliated with the Liberian Bible Translation and Literacy Organization (LIBTRALO) in Monrovia. While collecting the Swadesh list in Gbanquoi, a village in the Southern Mano zone, we were able to compare the variants and check for mutual intelligibility. In general, Pe Mamy, the speaker of the Maa dialect, had no much trouble understanding Liberian Mano, with the obvious exception of borrowings from English, a language he does not speak.

7In Guinea, many Mano are fluent in local majority languages, Kpelle and Maninka, while in Liberia, especially in the South of the Mano zone, many Mano speak Dan.

8In what follows, I will mostly focus on the comparison between the Maa dialect in Guinea and Southern Liberian dialect of Mano (SLM), which is the most distinctive of all the Liberian varieties if compared to Maa. Occasionally, I will provide some information on other dialects: Northern Liberian Mano (NLM), Central Liberian Mano (CLM), Kpeinson and Zaan.

Phonology

9According to preliminary observations, the segmental inventory in Guinean and Liberian varieties of Mano is identical. The inventories of the phonemes are given below in Tables 1 and 2. All Mano varieties have three tones whose functions are similar.

Table 1. Mano vowels

oral vowels			nasal vowels
i		u	ḭ		ṵ
e		o
ɛ		ɔ	ɛ̰		ɔ̰
	a			a̰

10In addition, Mano features a syllabic nasal /ŋ/.

Table 2. Mano consonants

	labial	alveolar	palatal	velar	velar labialized	labio‑velar
implosive	ɓ
plosives unvoiced	p	t		k	kw	k͡p
plosives voiced	b	d		g	gw	g͡b
fricatives unvoiced	f	s
fricatives voiced	v	z
sonants oral	w	l	j
sonants nasal	m	n	ɲ	ŋ	w̰

11The differences concern the suprasegmental level: assimilation consisting in vowel and consonant adjustment of suffixes and enclitics, processes of fusion in several morphologically independent lexemes, and differences in patterns of phonotactics resulting from different assimilation rules and fusion patterns. In addition, as I will show in Section 4 regarding lexical correspondences, SLM shows a tendency towards vowel shortening and consonant elision in the intervocalic position in non‑derived morphological units. At this point, it should be considered an idiosyncratic process, as no regular correspondences can be established, but it correlates with other, more regular tendencies in phonetics.

Assimilation

Vowel assimilation

12Liberian and Guinean varieties of Mano have different patterns of assimilation of vowels and consonants. In the Maa variety, the verbal suffixes –à (gerund, counterfactual) and –á (conditional, irrealis), as well as the demonstrative ā (which has free variants yā and yāā) and the topicalizer ā assimilate to the previous vowel. In the Liberian varieties, including SLM, assimilating markers that I tested are: the gerund suffix –à and the demonstrative marker / topicalizer ā, which also undergo assimilation, although the pattern is different from Maa. In what follows, I only discuss the patterning of gerund suffix and the demonstrative / topicalizer.

13In the Maa dialect, the vowel in the gerund marker ‑à obligatorily copies the nasality from the previous vowel, while the assimilation by the place of articulation is optional. The most common pattern in natural speech seems to be that the vowel in the gerund assimilates with semi‑open vowels (ɔ, ɛ, ɔ̰, ɛ̰), while the semi‑closed and closed vowels (o, e, u, i, ṵ, ḭ) typically do not trigger assimilation. As for the demonstrative marker, it assimilates by nasality. While assimilation by the place of articulation is accepted in elicitation, it never occurs in natural speech. The pattern of assimilation of the topicalizer ā, which is cognate with the demonstrative ā, is the same as for the gerund and the assimilation is frequent.

14In contrast, in SLM, the vowel in the gerund assimilates with all vowels, copying nasality and place of articulation. In addition, in case it follows a closed vowel, an additional variant is available: the vowel of the gerund copies the nasality and labialization features and becomes semi‑open, thus, after /i/ or /ḭ/, the vowel of the gerund becomes /ɛ/ or /ɛ̰/, respectively, and after /u/ or /ṵ/, it becomes /ɔ/ and /ɔ̰/, respectively. As for the demonstrative ­ā and the topicalizer in SLM, after all vowels but /a/ or /a̰/, the vowel of the demonstrative becomes semi‑open with the same labialization and nasalization pattern as the previous vowel. Some examples of contrasting vowel assimilation in Maa and in SLM are given in Table 3.

Table 3. Vowel assimilation in Maa and SLM

15To sum up, vowel assimilation in SLM is either full (in case of assimilation to open vowels and all vowels in the gerund form) or partial (in case of closed and semi‑closed vowels in all forms, where the assimilating vowel keeps the same quality of labialization and nasalization, but the degree of openness may be one or two steps removed: thus, the assimilating vowel becomes /ɛ/ when the preceding vowel is /e/ or /i/). In Maa, the assimilation is either full, or absent and only semi‑open vowels trigger assimilation by the place of articulation.

16In addition, according to the description provided by deZeeuw and Kruah, the assimilation of the demonstrative marker by the place of articulation is optional. When it happens, however, it is always partial: the vowel becomes semi‑open (deZeeuw & Kruah 1981: 114). Thus, the system described by deZeeuw and Kruah is intermediate between the one in Maa and the one in SLM. It correlates with my observation that their description is based on the Northern or Central Liberian variety spoken in the area right in between Maa and SLM. In the Bible translation, the demonstrative never assimilates by place of articulation, the topicalizer assimilates only to the semi‑open vowels, while the gerund marker always fully assimilates. Therefore, the system reflected in the Bible translation is also situated in between Maa and SLM, although it is different from the assimilation described by deZeeuw and Kruah and, presumably, it is closer to Maa.

Consonant assimilation

17In Mano, the syllabic nasal ŋ also triggers assimilation: while the nasal itself assimilates with the following consonant by place of articulation, the consonant adopts the nasality feature. The combination results in a nasal geminate. The most prone to assimilation is the combination of the first person singular non‑subject pronoun ŋ̄ and its syntactic head: be it a possessum in the inalienable possessive construction, a transitive verb, or a postposition.

Table 4. Consonant assimilation

Fusion

19In both Maa and SLM, fusion is common between several items. In Maa, fusion typically concerns pronouns and is different from assimilation in being irregular and in involving regressive, rather than progressive, assimilation: kɛ̄ à do 3sg → kāà; wɔ́ ī cop.neg 2sg → wɛ́í, wáí. It is rare that fusion would concern more than two morphological units, with the exception of the fusion of the gerund form with a postposition ká, which is so regular that it should be considered a special morphological form of the verb: dɔ̄‑à ká lay ger with → [dɔ̀á]. With the latter exception, typically Maa speakers do not use fused variants when speaking slowly.

20In SLM, many cases of fusion between two or more (morphologically independent) items were attested. In contrast with fusion in Maa and with assimilation patterns in both dialects involving specific morphological units, the fused combinations in SLM involve a broader spectrum of units. The above‑mentioned assimilation follows strict rules, while fusion described in the present section involves a wider set of processes, including regressive assimilation, but also consonant elision, and the exact rules are yet unknown. The native speakers are very aware of this fusion process, and speakers of different dialects repeatedly pointed to each other's differences in pronunciation. Although SLM speakers can decompose the fused item into separate ones for the purposes of explanation and glossing, the fused variant may be used even when pronounced very slowly.

21Some attested examples are presented in (1).

(1)	a.	yī ɛ̄ there top → [yɛ̄ ɛ̄]
	b.	óò ī 3pl.ipfv 2sg → [wéì]
	c.	sɔ̰́ɔ̰́ ká tooth with → [sɔ̰́ā̰]
	d.	ló mā ɓō go:foc 1sg.pst>3sg implement → [ló māā]
	e.	bḭ́ āà night 3sg.prf → [bḭ́ā̰à̰]
	f.	ŋ̄ wɔ̄ tá̰à̰ 1sg.pst lie ground → [ŋwà̰á̰à̰]
	g.	gà̰ mɔ̀ foot on → [gà̰ṵ̀]
	h.	ká lé mɔ̀ house mouth on → [kálɔ́ɔ̀, kálɔ́ṵ̀]

22The most common processes attested are: elision of the intervocalic consonant (1c, 1d, 1f—1h) and vowel assimilation, both in place of articulation (1a, 1b, 1d, 1f, 1h) and nasality (1c, 1e, 1f), and both regressive (1a, 1b, 1f, 1h) and progressive (1c, 1d, 1e). In 1b and 1c there is a decrease in the number of moras. Note that in both 1g and 1h the postposition mɔ̀ changed into a low‑tone syllabic nasal /ŋ̀/ (pronounced as [ṵ̀]).

23Fusion may cancel assimilation process. Thus, in the combination of the noun lēēnɛ̀fú ‘girl’ with a demonstrative the demonstrative is expected to assimilate and become [ɔ] (see the last example in Table 3). In a combination with a 3sg pronoun à, however, the assimilation did not occur, and the overall combination lēēnɛ̀fú ā à child.foc dem 3sg was pronounced as [lēēnɛ̀fúāà].

24As my analysis of the lexical correspondences will show (Section 4), the same features, elision of consonants and the decrease in the number of moras, characterizes some SLM reflexes in comparison with their cognates in other dialects.

Phonotactics

25As a result of the fusion process, new syllable types are attested in SLM that are not attested in Maa. In particular, the abundance of CVVV structures is observed (cf. [bḭ́ā̰à̰], [ŋwà̰á̰à̰]). These units could be analyzed as combinations of two syllable structures, CV and VV or CVV and V, but thanks to the assimilation process they could be considered single metrical feet (Vydrin 2010).

26Because of the elision of intervocalic consonants and assimilation patterns SLM features original combinations of vowels, such as combinations of semi‑closed and semi‑open vowels: kpɛ̀ē ‘to dawn’ (cf. Maa kpàlē), kéɛ̄ ‘this’ (cf. Maa kéā).

27An interesting feature of Mano phonotactics is that combinations of back and front vowels, which are generally prohibited, are nevertheless possible after an (optionally) labialized velar consonant: cf., in Maa, k^(w)ɔ̀ī ‘firewood’, k^(w)ōí ‘behind’, k^(w)ɔ̀nɛ̀ ‘eggplant’. In the CVV combinations of this type, V₂ is usually /i/, with one exception: k^(w)ɔ̀ɛ́ ‘near’. In fast speech such combinations are pronounced with both front vowels and a strong labialization: [k^wɛ̀ī] ‘firewood’, [k^wēí] ‘behind’, [k^wɛ̀nɛ̀] ‘eggplant’, [k^wɛ̀ɛ́] ‘near’. In SLM, the default variant is with a back – front combination and the labialization of the velar consonant is much weaker ([kɔ̀ī] ‘firewood’). Note that the native linguistic intuition strongly confirms the possibility of the back – front vowel combination, and also the optionality of labialization, which is reflected in the spelling of a Liberian village: Gbanquoi or Gbankoi. Although front‑back combinations are rare in SLM, they are not prohibited, in contrast with Maa: cf. sùē 'fingernail’ (SLM) vs sèē (Maa).

Morphosyntax

28The morphosyntax of all Mano dialects seems largely similar, although it may differ in minor details. I have collected several oral narratives in the SLM dialect, and in particular, the narrative chain and the complex clauses that occurred there (temporal and relative) were formed with exactly the same model as in Maa. Three features stand out: first, differences in the system of demonstrative markers. Second, the SLM dialect has a different stem for the third person sg. pronominal marker: ɛ, instead of e in all other Mano varieties for which data are available. And the third feature concerns a distinction in the pronominal paradigm concerning the first person non‑singular pronominal stems.

Demonstrative system

29The Maa dialect has five demonstratives: tɔ́ɔ̄, dḭ̀ā̰, wɛ̄ ~ ɓɛ̄, yā ~ ā ~ yāā (there are also variants assimilated by place of articulation and nasality to the previous vowel, see Section 2.1) and kílīā ~ kílīɓɛ̄. Tɔ́ɔ̄ and dḭ̀ā̰ are proximal and distal demonstratives, respectively, and are typically used to draw attention to discourse new referents in the interactive scene. Kílíā ~ kílíɓɛ̄ are used as anaphoric markers. Although in discussions speakers interpret wɛ̄ as more proximal than yā (and translate them by French terms celui‑ci ‘this’ and celui‑là ‘that’, respectively), the two markers are largely interchangeable in discourse. When used adnominally, wɛ̄ and yā cover all the demonstrative functions suggested by Himmelmann (1996): they are both used in exophoric reference (marking referents present at the interactive scene), but they are especially common in endophoric functions, marking referents not present at the interactive scene, including discourse‑referential, anaphoric, recognitional, but also cataphoric functions. In addition, the marker kéā, in functions similar to wɛ̄ and yā, is used by some speakers of Maa and is typical for the speakers of Kpeinson.

30In Liberian Mano, the visible demonstratives tɔ́ɔ̄ and dḭ̀ā̰ were not attested (although the reason could be that in Maa, they are typically used in conversations which were not recorded among the SLM speakers). In elicitation, the kéɛ̄ marker was suggested as a proximal marker, ā as distal and kílíɛ̄ (cognate of Guinean kílíā) as anaphoric, wɛ̄ was not mentioned even if I tried to directly elicit it. In the recorded texts, however, the wɛ̄ marker was used with highly prominent referents (especially in the recognitional function, where the referent is cognitively available without a prior mention), while ā was systematically used in the anaphoric function, including in the bridging context where the referent was introduced by some contextual relation.

31In the following excerpt from a narrative in the Zaan dialect, Guinea, the demonstrative wɛ̄ was used in the anaphoric function introducing the third mention of the noun nínà ‘devil’.

32(2) 1. wáà gèè à lɛ̀ɛ̄ kélɛ̀ “nínà āà ī nā sí, āà ló á ká.”

33‘They say to him: a devil took your wife, he carried her away.’

342. lɛ̀ tɛ́ nínà ē ló à nā ká yí ā làá dɔ̄, áà ē ɲɛ̀ɛ̄ kpɔ̀ à ká, áà ē ɲɛ̀ɛ̄ kpɔ̀ à ká, làá gɛ̰̀.

35‘He does not know the place where the devil carried his wife, he looks for it, he looks for it, he cannot find.’

3.	ló	tɛ́	nínà	wɛ̄	ā	ɓō	à	nā	ká	ā
	go	foc	demon	dem	3sg.pst>3sg	take.off	3sg	wife	with	top

ā	yà	yílí	gáná	dò	yí.
3sg.pst>3sg	sit	tree	root	indef	in

36‘The devil having carried away his wife, he put her in the root of a tree.’

37The following excerpt from a story told in the SLM dialect speaks about a group of young men and women and in particular about a woman among them who had no teeth. They went bathing in the water hole, and while they were doing that, a bird stole their clothes. The demonstrative ā (including its assimilated variant ɛ̄, see above) is used in the anaphoric function with nɔ́ɔ̄ɓé ‘the children, the youngsters’ and with mī ‘person’, referring to the woman. Wɛ̄, in contrast, is used with the pronoun àyē ‘him/her’ referring, again, to the woman. The referent was already re‑introduced by the noun mī 'person’, so wɛ̄ does not serve the anaphoric function but is used to emphasize the prominence and topicality. In addition, wɛ̄ is used with the noun sɔ̄ ‘cloths’. Although the noun was introduced in the prior narrative, here the referent is different: the woman talks about her own cloths, and it is the first time she realizes that they are gone – therefore, it is, again, the prominence of the referent that determines the choice of the wɛ̄ marker.

(3)	nɔ́ɔ̄ɓé	ɛ̄,	wɛ̀lɛ̀	wā	ɓō	yíí	bà	ā,
	child.pl	dem	stand.up	3pl.pst>3sg	implement	water	in	top

“ŋ̀	sɔ́	wɛ̄	lɛ̄	mɛ́?”
1sg.poss	cloth.foc	dem	3sg.exi	where

38‘The kids, when they got out of the water, then that person, she said: my clothes (lit.: cloth), where are they?’

(4)	1.	Sèé	ē	ló	kàā	dà‑pìà	yíí	píé.
		p.n.	3sg.pst	go	hook	throw‑inf	water	at

40‘Se went to fish at the river (lit.: Se went throwing hook).’

	2.	Sèé	ē	mámá	wàà	kàā	mɔ̀.
		p.n.	3sg.pst	bait	enter	hook	on

41‘Se put bait on the hook.’

	3.	Sèé	ē	kàā	ā	dùò	yíā	bà.
		p.n.	3sg.pst	hook	dem	throw	water.dem	in

42‘Se threw the hook in the river.’

43Note that such usage of ā is very unlikely to be a calque from English: English provides no model for the equally systematic avoidance of the anaphoric marking with the dependents of postpositions, as in ex. 4.2.

Third person sg. segmental base

44All Mano dialects have a rich inventory of pronouns which employs a set of one to four segmental bases. Pronouns are often similar in segmental structure and differ in tone or vowel length. In the SLM dialect, the segmental base of the third person sg. pronouns is ɛ, rather than e, as in all other dialects, Guinean and Liberian.

(5)	ɛ̄	ɛ̄	nā	gɛ̰̀	lɔ́ɔ́í	(SLM)
	ē	ē	nā	gɛ̰̀	lɔ́ɔ́í	(all other dialects)
	3sg.pst	3sg.refl	wife	see	market

45‘He saw his wife at the market.’

46Although in the literacy materials and in the Bible translation only the e variant is attested, deZeeuw and Kruah give both variants for the past series (1981: 128) and for the reflexive.

First person non‑singular pronouns

47The manual of Liberian Mano by deZeeuw and Kruah (1981) gives two pronominal stems for the first person non‑singular pronouns. According to them, the opposition is based on clusivity: the ko stem is exclusive and the kɔa stem is inclusive.

(6)	kò	ɓī
	1pl.ip	2sg.emph

49‘you (sg.) and I (lit.: we including you (sg.))’

50‘you and I and some other people’

51In natural speech, however, the kò pronoun can be used to refer to more than two participants:

(8)	kò	mɛ̄ŋ́	mìà	nù
	1pl.ip	Maninka	person.pl:cs	pl

52‘we and the Maninka’

53Therefore, it is possible that the kwa stem is used with three or more participants and the ko stem can be used with two and more participants. More data is needed to specify the semantics of the pronouns.

54In the Maa dialect of Guinea, the same pronominal form is used in both contexts.

(9)	kò	ɓī
	1pl.ip	2sg.emph

55‘you (sg.) and I’

(10)	kò	ɓī	nì
	1pl.ip	2sg	pl

56‘you and I and some other people’

57In addition, the Maa dialect (and, probably, other Guinean dialects as well) has the kɔa stem for the first person plural pronouns which is used in specific series of pronominal auxiliaries, such as perfect (kɔ̄āà ‘1pl.prf’) or prohibitive (kɔ́áá ‘1pl.proh’), as well as in certain portemanteau series, fused with the third person sg pronoun à (kō ‘1pl.exi’ vs kɔ̄ā ‘1pl.exi>3sg’). According to my notes, in the Kpeinson dialect of Guinea, which is the closest to the Liberian dialects, the kwa stem is a free variant and can replace both ko and kɔa. Information on full pronominal inventories in SLM is not yet available.

Lexicon

58The basic lexicon (at least, within the limits of the 100‑word Swadesh list) is largely the same in all Mano dialects that I have data on. In most cases, it is clear that we are dealing with the reflexes of the same proto‑lexemes. In SLM, however, generally due to the processes of consonant elision and vowel shortening, some reflexes look different (cf. in Maa, nānà ‘tongue’ and in SLM, nāà; in Maa, lēē ‘woman’ and in SLM, lē). The lexical item that shows the most variation is the word ‘fingernail’. In Zaan, in is sòō, in Maa, sèē, in Kpeinson, sìē, in SLM, sùē. In several lexemes, the Kpeinson dialect differs from other Guinean dialects and is closer to the Liberian varieties. Whenever CLM is different from SLM, it is closer to the Guinean varieties. Therefore, when it comes to lexical correspondences, there is a clear dialectal continuum.

59Rare are the cases where a cognate of a lexical item in one dialect cannot be found in another dialect. Such cases are: tííkpé ‘small’, dɔ̀kɛ̄ ‘give’, tɔ́ɔ̄ ‘this, visible’ in Maa, which do not exist in Liberian varieties, where ɓéŋ́, péŋ́ are used for ‘small’, gbā and nɔ̄ for ‘give’ and where dedicated visible proximal demonstratives are not attested. There are also two expressions, sɔ̰́ɔ̰́ dɔ̄ (lit. ‘tooth place’) ‘to bite’ in Maa and CLM which corresponds to kṵ́ sɔ̰́ā̰ (lit. ‘catch tooth.with’) in SLM, and yíí là kɛ̄ (lit. ‘water surface do’) 'to swim’ in Maa, which corresponds to yíí kɛ̄ in SLM.

61Table 5 below summarizes some of the differences in the lexicon (including different reflexes of the same cognate, as well as different stems, which are marked by bold characters) between the dialects. The first column gives a common English translation, the differences in semantics are given in the respective fields. The second column gives correspondences in the Maa dialect of Guinea, the second column gives correspondences in the Southern Liberian Mano with some additional comments about Central Liberian Mano or Northern Liberian Mano; the latter were obtained from a speaker of Central Liberian Mano, there is no first‑hand data. Finally, the last column gives correspondences found in the language manuals with marked source page number; by default, the data is taken from (deZeeuw & Kruah 1981), whenever it was taken from (Neal et al. 1946), it is explicitly marked.

Table 5. Lexical correspondences between Mano dialects

Interdialectal variation motivated by different contact situations

62In this section, I will discuss a hypothesis that some of the dialectal differences in Mano could be explained by different contact situations: in the North, and especially in Guinea, Mano is influenced by Kpelle, while in the South of Liberia, it is influenced by Dan. In Section 5.1, I discuss phonological variation, while in Section 5.2 I present lexical influence. In Section 5.3 I present the limits of contact explanation.

Phonological influence

63One of the discussed points of variation is the realization of the combinations of a velar consonant with a back vowel, followed by a front vowel. Thus, in Maa, k^(w)ɔ̀ī ‘firewood’ in fast speech is pronounced [k^wɛ̀ī]. Note the spelling of this lexeme attested in deZeeuw and Kruah (1981: 139) is kwàī. In Guinean Kpelle, the same patterns of alternation are attested: kɔ̀lɛ̀ [k^wɛ̀lɛ̀] ‛near’, kóníŋ̀ [k^wéníŋ̀] ‛scratch’ (Konoshenko 2017: 286). In Liberian Kpelle, the dictionary entry is kɔlɛ ‘vicinity’ (Leidenfrost & McKay 2007:138). In SLM, the default variant seems to be with a back – front combination, whereas the labialization of the velar consonant is much weaker ([kɔ̀ī] ‘firewood’). Therefore, the alternation in Guinean Mano and lack of it in Liberian Mano could be due to Kpelle influence or shared patterns in phonetics.

64Another point of comparison are the patterns of consonant assimilation in the context of the syllabic nasal. In Liberia, only /ɓ/, /l/, /y/, and /w/ can be nasalized, while in Guinea, in addition to these four consonants, assimilation of /d/, /s/, /z/, /k/, /g/ and even /gb/ is attested. A possible explanation could be the influence of the Kpelle system of consonant alternation.

65In Kpelle, certain phonological, as well as morphological contexts trigger alternation of the initial consonant. The alternating pairs are the following: p → b, t → d, k → g, kw → gw, kp → gb, hw → v, h → z, ɓ → m, l → n, y → ɲ, w → ŋw. The last four pairs are directly parallel to the assimilated and non‑assimilated consonants /ɓ/  → [mm], /l/  → [nn], /y/  → [ɲɲ], and /w/  → [ŋ^wŋ^w] in all Mano dialects, with the difference that in Mano, in contrast to Kpelle, the result of assimilation is a geminate. Consonant alternation in Kpelle can be triggered by morphonological processes, including a combination, at the deep level, with the high‑tone nasal prefix expressing a first‑person singular pronoun of the polyfunctional series. In Mano, the assimilation is also triggered by a combination with a first‑person singular pronoun. Therefore, consonant alternation in Kpelle and consonant assimilation in Mano can occur in similar morphosyntactic contexts and can give, at least for certain consonants, very similar results:

66/ŋ́ + lèě/ → [ńéè] ‘my mother’ (Kpelle)

67/ŋ̄ + lèē/ → [n̄n̄èē] ‘my mother’ (Mano)

68The results of assimilation in Mano, however, do not always correspond to the results of alternation in Kpelle (/k/ in Kpelle becomes [g], but in Mano it becomes [ŋ]) and the contexts of assimilation and alternation do not coincide (in Kpelle, the third person singular polyfunctional marker also triggers assimilation). Crucially, consonant alternation in Kpelle occurs in a much wider set of consonants than regular assimilations in Mano. One could assume that Mano has taken the regular system of consonant alternation in Kpelle as a model and extended, albeit irregularly, the set of possible consonants that can undergo assimilation to include some new elements, such as dental consonants.

70In contrast, Dan provides a good model for different kinds of fusion processes: zı̰̄ȁ̰ȁ̰ = zı̰̄ā̰ā̰ tȁ ‛on the road’ (Vydrin 2017: 478); cf. in SLM ŋ̄ wɔ̄ tá̰à̰ 1sg.pst lie ground → [ŋwà̰á̰à̰]). In Dan, postpositions regularly fuse with the nouns they govern, giving rise to an emergent system of cases (Vydrin 2011). In particular, Dan‑Gweetaa features an instrumental case: gɛ̰̏ ‛leg’, gɛ̰̄ɛ̰̄ ‘leg.instr’ (Vydrin 2017: 485), cf. in SLM sɔ̰́ɔ̰́ ká tooth with --> [sɔ̰́ā̰].

Kpelle and Manding lexical influence

71In Guinea and northern Liberia Mano has long been in close contact with Kpelle, a Southwestern Mande language, through warfare (both legends about foundations of Guinean Mano villages that I collected contain reference to wars with Kpelle), trade, and intermarriage. The latter is especially common in large multilingual towns. My primary language consultant and his elder sister, both residing in Nzerekore, the capital of the region, are both married to Kpelle. While many Mano speak Kpelle, there are much fewer Kpelle who speak Mano, which could be due to the fact that Kpelle outnumber Mano. In addition to the long‑term symmetric contact, Kpelle has served as a dominant language in education and other domains: many official names of Mano villages are Kpelle by origin; in the early years of Guinean independence the primary education in the Mano – Kpelle zone was organized in Kpelle, and Mano was officially regarded as a dialect of Kpelle. The superstratal influence is especially salient in the Roman Catholic Church, where Kpelle is the language of the ecclesiastic authority and much of the religious texts are translated from Kpelle. Liberian Mano, especially in the North, is also influenced by, or shares common innovations or retentions with, Liberian Kpelle. However, I have much less data on the vocabulary of Liberian Mano and on the sociolinguistic situation to provide any convincing argument of the contact situation.

73Table 6 gives some examples of Mano – Kpelle correspondences, while Table 7 adds Manding to the picture. The data from Liberian Mano (SLM and CLM) is contrasted with the data from Liberian Kpelle (Leidenfrost & McKay 2007), from the Maa dialect in Guinea, from Guinean Kpelle (Konoshenko), (Leger 1975) and, whenever there is a correspondence with Kpelle or Mano, from Maninka, a Manding variety with which Guinean Mano and Kpelle are in contact (Vydrin n.d.).

Table 6. Mano – Kpelle correspondences

	Liberian Mano	Liberian Kpelle	Guinean Mano	Guinean Kpelle
no	kpàò	kpa	kpàò	kpàà
but	kɛ̄ɛ̄	kɛ́ɛ, kɛ́lɛ	kɛ̄ɛ̄, kálá	kɛ́láá
to thank, thank you	zúò	sɛɣɛ̂, máma	zúò, sɛ̀kɛ̄ɛ̀, mámá	hɛ̀ɣɛ́ɛ̀, mámá
judge(ment)	mɛ̀ŋ́sà (judgement)	mɛni saa affair cut (to judge)	mɛ́ŋ́sà (judgement)	mɛni kiti teɣe affair judgement cut (to judge)
sin	sɔ̀ŋ̄ yɔ̄ɔ̄	sɔŋ nyɔŋ bad behavior	náá, náŋá (= ‘malediction')	láŋáŋ́
food	léɓèlè	kɔ́nɔŋ	kɔ́nɔ́	kɔ́nɔ́ŋ́
to read	lònú; lònɔ́	lóno	gèē (= ‘say')	lónóŋ; lɔ́nɔ́ŋ
small	ɓéŋ́, péŋ́	keni	ɓéŋ́, péŋ́, tííkpé	tííkpé

74Table 6 presents some lexical correspondences between Mano and Kpelle without any Manding influence. It shows that there are some common features in the Mano – Kpelle lexicon in both countries, such as the interjection ‘no’, as well as some clustering of Guinean and Liberian varieties together (as in the case of ‘sin’ or ‘small’). Quite frequently, however, we observe the same term in Guinean and Liberian Kpelle which gets transferred to only one Mano variety: thus, lònú; lònɔ́, the Liberian Mano equivalent for ‘to read’ is likely a borrowing from Kpelle lóno, in contrast to Guinean Mano which uses the verb gèē ‘to say’ in this function. Note that both Guinean and Liberian Mano words for ‘judgement’ come from a Liberian Kpelle expression ‘to judge’, while Guinean Kpelle uses a different (although related) expression. The conjunction ‘but’ shows interesting patterns: one of the Guinean Mano variants, kɛ̄ɛ̄, is also used in Liberian Mano and Kpelle, while another Guinean Mano variant is used in both Kpelle varieties. Finally, the Guinean Mano word for ‘food’ is used in both Kpelle varieties, while Liberian Mano uses a different word. The most striking example, however, is the ‘thank you’ equivalent: while Liberian Mano keeps what must be an original Mano expression, which also occurs in Guinean Mano, but only with reference to God, Guinean Mano uses the expressions common to Kpelle (and also, as noted in Table 7, a Manding borrowing via Kpelle). These example show patterns of mutual influence of Guinean and Liberian Mano and Kpelle varieties, as well as influence across the political borders, although it seems that the influence of Kpelle on Mano is somewhat stronger in Guinea than in Liberia.

75Another observation is that /h/ in Guinean Kpelle regularly corresponds to /s/ in Guinean Mano. Mano lacks /h/ phoneme, so a correspondence had to be found. A plausible reason why /s/ was chosen as a correspondence to /h/ are contacts across the political border which lead to an awareness that /h/ in Guinean Kpelle corresponds to /s/ in Liberian Kpelle (and, thus, may correspond to /s/ in Mano). Such awareness is especially important in those cases where a Kpelle borrowing is present only in Guinea, such as in the case of hɛ̀ɣɛ́ɛ̀ ‘thank (you)’ in Guinean Kpelle which corresponds to sɛɣɛ̂ in Liberian Kpelle and sɛ̀kɛ̄ɛ̀ in Guinean Mano. In addition, /h/ could be an innovation: a proto‑phoneme in Kpelle could be *s, and consequently, early Kpelle borrowings into Mano could be introduced with /s/, while recent borrowings could be assimilated following the model of earlier borrowings and common Mande retentions. The same /h/ – /s/ correspondence will characterize the borrowings with Manding origin to which I shall now turn.

Table 7. Lexical correspondences between Mano, Kpelle and Manding

	Liberian Mano	Liberian Kpelle	Guinean Mano	Guinean Kpelle	Manding
thousand		wála	wáá	wáá	wáa
price		sɔ̰ɔ	sɔ̄ŋ́	hɔ̀ŋɔ̌	sɔ̀nkɔ́
money		sɛŋ kao, kâpa	wɛ́lì	wálî	wádi
soap		saafûlɛ	sàfná	hahvunɛ	sàfína (Ar. ṣābūn)
sacrifice		sala	sálà	hɛ́láà	sáraka (Ar. sadaqa, (voluntary) offering)
teacher	kīīzɔ̰̀ɔ̰̀mì book show person	kɔlɔ lɛ nuu book show person, kaamɔ̂ɔ	kàlàmɔ́ɔ̀	kalamô	kàramɔgɔ
pray, prayer	sɛ̀nɛ̄ ɓō	sɛ̂li ɓo (Muslim prayer)	sɛ̀nɛ̄ ɓō	hɛ́lî	séli (Ar. çalaa)
book	kīī (= ‘skin')	kɔlɔ (= ‘skin')	sɛ́ɓɛ̀	hɛ́bɛ̂	sɛ́bɛ (Ar. çafḥa ‘page’)
to thank, thank you,		ɓáliká	bálíká	bálíká	bárika
church, Christianity	sɔ́ɔ́ī < Eng. church	derivates from Kôrai ‘Christ’ or ɣâla ‘God’	kānà	kàláŋ̀	< kàrán ‘read, teach’ (Ar. qara’ ‘to read’, cf. Quran)
heaven	lɛ̀í, wálà pà God at	ɣâla‑taa God at	ārzānā, wálà pà God at	árzáná	àlijana (Ar. al jannah)
save	lā		kísí ɓō, lā	kihiɓo	kísi
judgement			kítí	kítí	kíti (Ar. qaḍiyya)
insult, offense			ɓàkà ɓō	ɓaɣa ɓo	baga

76Table 7 provides some examples of borrowings from Manding to Kpelle and (then) to Mano. A domain that seems to be very influenced by Maninka is trade: note the words wáa ‘thousand’ and sɔ̀nkɔ́ ‘price’ borrowed from Manding in Guinean and Liberian Mano and Kpelle. The words for commercial goods, such as ‘soap’, were also borrowed and can ultimately be traced to Arabic (note that English or French words for soap are also Arabic borrowings). Note that the word for ‘money’ was borrowed in both Guinean varieties, but not in Liberian Kpelle (I do not have evidence for Liberian Mano). Other cultural lexica were borrowed in Guinean varieties, but to a lesser extent in Liberia: the examples are the equivalents for ‘book’, ‘teacher’, ‘to thank’. The same concerns some basic religious terminology, such as ‘sacrifice’ or ‘prayer’: note that in the dictionary of Liberian Kpelle a Manding borrowing has an unambiguous connotation ‘Muslim prayer’. However, when it comes to other religious vocabulary in Liberia, native terms are strongly preferred, while in Guinea Manding borrowings, many of which are Arabic by origin, abound.

77Although the comparative lists in Tables 6 and 7 are small, the general impression is that there is less leakage across the political border of the Maninka‑influenced cultural lexicon than of basic lexical items shared by some Kpelle and Mano varieties. The notable exception is the trade lexicon and some basic religious vocabulary which is shared across the region. This suggests that extensive cultural and religious influence of Manding is more recent than the political borders. This is also supported by the history of the Christianization of Kpelle and Mano, which was independent in both regions and started after the political borders between Guinea and Liberia were traced. Crucially, in Guinea the Catholic missionaries relied a lot on Manding‑speaking interpreters in search for the equivalents in Kpelle, which explains the great number of Manding borrowings in religious vocabulary (Lelong 1949). The Christianization of Mano started much later than the Christianization of Kpelle and missionaries working among Mano heavily relied on Kpelle as a model, which explains how Manding borrowings further spread to Mano.

78Note also that with a couple of exceptions in Guinea and Liberia Mano and Kpelle cluster together in their borrowing / retention patterns, as well as patterns of polysemy: both in Liberian Mano and Liberian Kpelle the words kīī and kɔlɔ mean ‘skin’ and ‘book’, while in Guinean Mano and Guinean Kpelle they only mean ‘skin’; the Guinean Mano word sɛ́ɓɛ̀ ‘book’ is likely to be a borrowing from Guinean Kpelle hɛ́bɛ̂ (which is a borrowing from Manding).

79In summary, the patterns in the spread of common vocabulary show a complex multilayered picture of horizontal contact between Mano and Kpelle and across different dialects, of old contact with Manding via trade and, to a lesser extent, religion, and a more recent introduction of Manding borrowings intensified (or motivated) by the translation practices of the Guinean missionaries and limited by political borders.

80By contrast to Guinean Mano, the lexicon of SLM could be influenced by Dan. ‘To bite’ in Guinean Mano (and in CLM) is sɔ̰́ɔ̰́ dɔ̄ ‘tooth place’, while in SLM, it is kṵ́ sɔ̰́ā̰ ‘catch tooth+with’. In Dan, the model is the same as in SLM: kṵ́ sɔ̰̋ɔ̰̋ ‘catch tooth.instr’(Vydrine & Kességbeu 2008:22). The same applies to yíí là kɛ̄ (lit. ‘water on do’) ‘to swim’ in Guinean Mano and yíí kɛ̄ (lit. ‘water do’) ‘to swim’ in SLM: in Dan, ‘to swim’ is yi̋ kʌ̄ (lit. ‘water do’) (Vydrine & Kességbeu 2008:295). Dan, just like SLM, has only one marker of plurality, which is nȕ (cf. nù in SLM vs vɔ̀, additive plurality, and nì, non‑additive, including associative, plurality in other Mano varieties). More information is needed to account for SLM patterns of borrowing / retention.

The limits of contact explanation

81The issue of contact does not obviously explain all the interdialectal differences. In particular, if in Guinean Kpelle the rules of assimilation of the gerund are almost identical to those in SLM (Konoshenko 2017: 319), why should Guinean Mano be different? A different example: if both Kpelle and Dan have inclusive vs exclusive distinction in pronouns, why did Guinean Mano lose it, and Liberian Mano attach a different value to the different pronominal forms in the paradigm?

82In addition, there are some features common to Mano, Kpelle and Dan, such as agreement patterns studied by Konoshenko (2015: 176‑177), which could be due to some larger areal influence.

83As noted in Section 1, Mano varieties spoken to the south of Gbanquoi are expected to manifest more differences with respect to other Mano dialects. A comprehensive documentation of these and other Liberian dialects should be an object of a future study.

Abbreviations

84cs – construct state

85dem – demonstrative

86emph – emphatic

87exi – existential

88foc – focus

89indef – indefinite

90inf – infinitive

91ip – inclusory pronoun

92ipfv – imperfect

93pl – plural

94poss – possessive

95prf – perfect

96pst – past

97refl – reflexive

98sg – singular

99top – topicalizer