Memoire Online - The morphosyntax of ghomala' verbs: focus on inherent complément verbs and serial verb construction

In this chapter, I have discussed the morphosyntactic properties of Gh?maìlaì' verbs. In doing so, verbs have been classified following three criteria, namely the tone pattern, the syllable structure and the morphology of the verb. Looking at the verb structure, two derivative morphemes are attested in the language. These morphemes encode different semantic interpretations (reflexivity, reciprocity, attenuative...etc) and may affect the valency of the verb. The infinitive form, being

the nominalized form of the verb, can occupy different grammatical positions. As far as transitivity is concerned, one place predicates, two-place predicates as well as three place predicates are attested in the language. Some verbs can undergo transitive alternation. The work has argued for a three way distinction of intransitive verbs in Gh?maìlaì'. The chapter also addressed the issue of theta role assignment within transitive and the semantic relationship between the verb and its complements. It has been argued that most of the transitive verbs in Gh?maìlaì'^ì can select a generic-meaning DP as object. However, there are some verbs which do not take generic objects and require a more specific object. These ones have been referred to as Inherent Complement Verbs. The following chapter is devoted to the analysis of these verbs in Gh?maìlaì'.

Introduction

The previous chapter has discussed the categorization of Gh?maìlaì' verbs based both on their structural properties and their argument structure. It has been argued that argument structure depends both on the idiosyncratic properties of the verb as specified in the lexicon and the environment surrounding the verb. It has been shown that for some verbs, only specific DPs are required in order to give a particular meaning of the verb. These are Inherent Complement Verbs (ICVs), the raison d'être of the present chapter. Their specific complements are referred to as Inherent Complements (ICs) in the literature. This chapter aims at investigating their structural as well as morphosyntactic properties in order to see whether they are syntactically different from regular verbs or otherwise. In doing so, section one tackles the nature of ICVs and ICs by characterizing them. Their morphosyntactic properties are addressed in section two by looking at their behavior when they are used within some constructions in order to distinguish them from regular verbs. Section three tackles the derivation of ICV by laying emphasis on their argument structure.

This section aims at answering to the following question what is an ICV and IC? Although this issue has been briefly discussed in chapter one, some formal characteristics which help in identifying a verb as an ICV or a complement as an IC are addressed here. This is relevant in discussing some issues namely, whether the verb has any meaning contribution in the [verb_ noun] complex, whether the inherent complement is an argument of the verb and what is the right argument structure analysis of ICV constructions given the syntax-semantics mismatches they exhibit.

Nwachukwu (1987:22) defines an inherent complement verb as a verb «whose citation form is obligatorily followed by a meaning-specifying noun complement.» Thus, as claimed by Korsah (2011), an ICV is a verb whose function as predicate mostly depends on its complement.

The meaning of these verbs is tied to their complement as shown below, I gloss the verb as «Vx» in which x encodes an approximate meaning:

b. Ba^ka?m haì mku? faìlA
Bakam.PRS1 _Vgive respect priest «Bakam has honoured the priest»

It can be observed in (1) and (2) that the meaning of the verbs vary depending on the following element. For example, j?ì is interpreted as «suffer» in (1a) whereas in (1b) it has an interpretation of «dream». We also noticed that the Inherent complement occupies the structural position of an internal argument and the verbs are either translated as intransitive verbs (1) or as transitive verbs (2) in English. For this reason, some scholars regard constructions with similar syntactic and semantic structures as (1) and (2) in other languages like e.g. Igbo (Nwachukwu 1985, 1987) and Fon (Avolonto 1995) as intransitive and transitive constructions respectively.

The verbhood of the ICV itself based on its morphological properties is not dubious. The ICV inflects for tense, negation and aspect as it is illustrated in the following examples:

(3) a. Ba?ka^m kA-tâ-wâ d35 3èm pâ
Bakam PST2-NEG-PROG _Vsee dream NEG «Bakam was not dreaming»

We can observe in (3) that ICV inflects for both aspect and negation (3a) as well as future tense

Looking at ICVs in Gh?maìlaì', one may observe that some verbs can select variable complement to yield different meanings as shown in (1) above. This kind are referred to as regular ICVs (Korsah 2011) in the literature. Some ICVs are used with particular complements only. They are referred to as irregular ICVs. In this vein, Gh?maìlaì' ICVs can be grouped into regular and irregular ICVs based on their ability to take variable ICs to yield different meanings.

This kind of ICV is attested in the literature (Nwachukwu (1985 and 1987), Essegbey (1999), Korsah (2011 and 2013)). These ICVs select, in most cases, different complements to yield different semantics as shown in (1) and (2) above. The following are some examples from Igbo and Ewe:

Regular Gh?maìlaì' ICVs may have homophonous non-ICV counterparts in the language as shown in the following examples:

As outlined by the data in (7and 8) above, most of the verbs occurring in ICVs constructions also function independently as lexical verbs. No surface difference immediately sets the ICV in (7a) apart from the lexical usage in (7b). In both situations, the verbs combine with a noun phrase complement. What distinguishes non-ICV verbs from ICV ones is their ability to select different complements and maintain same meaning. They can take generic-complement, something which is impossible for ICVs.

According to Uchukwu (2004), ICVs that select different ICs to derive variant meanings form a cluster. Within the cluster, the verb roots possess a systematic meaning which is derivable from the meanings of all [ICV+IC] that have the same verb. The following clusters have been identified in the language under study:

It is possible to associate each ICV+IC in each of the above cluster with some underlying meaning.

For example, the j?ì cluster seem to have the meaning of «see» underlying in the various

combinations (see data in (1) above). The same situation is observable in the haì cluster wherein ICVs commonly have the meaning of «give» as shown in (2). Those in the taìm cluster seem to encode «move».

According to Korsah (2011), this kind of ICVs behave differently from regular ICVs in the sense that such verbs seem to depend on their complements in terms of meaning. In fact, irregular ICVs are used only with particular complements. They do not co-occur with other complements to yield different predicate meanings. The following ICVs have been identified as being part of this group:

Nâ- kx?Ì d?? INF-Vrun race

Nâ-tsà mti
INF-Vrelease saliva Nâ- t fw5 sî INF-Vsit ground

«to run» Nâ-?5 nwà

INF-Vknow matter

«to spit» Nâ-kwiì tÈ

INF-Vtake part

« to sit» Nâ-bv?Ì sî

INF-fall ground

«to be smart» «to defend» «to collapse»

The irregular ICVs may also have full lexical counterparts in the language. In this vein, they can co-occur with a generic-meaning complement as shown in (11) below:

(11) a. Ba?ka^m kà- wâ sÉ nwà
Bakam PST2-PROG _Vcount matter «Bakam was thinking»

b. Ba?ka^m k?Ì-wâ sÉ jwâ
Bakam PST2-PROG count thing «Bakam was counting thing»

As far as the semantics of ICVs is concerned, there are divergent point of views, in the literature, on whether the verb itself possesses a semantic content that contributes to the interpretation of the sequence or not. These disagreements are due to the fact that the regular word-for-word translation/ interpretation of ICVs constructions into other languages may break down. This breakdown stems from the fact that the ICV must occur with the inherent complement in

order to give a particular meaning. Thus, an ICV and its IC seem to be semantically bound such that if the verb is cited without its IC in the same construction, its meaning might be difficult to determine.

For some authors such as Avolonto (1995), reported by Essegbey (1999), the verb root in Fon ICV constructions is a verbalizer since the semantic content of the predicate is supplied by the inherent complement. This entails that the verbal part of the ICV construction in Fon has the function of turning the inherent complement into a verb. This rationale cannot hold in Gh?maìlaì' because this language has three class-changing processes which are reduplication, compounding and deverbatives (Moguo 2016). Furthermore, there is no suffix such as English ones -ise /-ize that changes words from other classes into verbs. The sole morpheme that occurs in front of the verb root in Gh?maìlaì' is the infinitive particle né.

The idea that ICVs are verbalizers is not plausible in Gh?maìlaì'. ICVs are always morpho-syntactically free. Furthermore, an ICV always precedes the IC and class-changing derivational morphemes are not attested in the language.

According to Nwachukwu (1987:40), the verb root lacks meaning without the inherent complement. As reported by Korsah (2011), this opinion is motivated by the fact that the regular one-to-one glossing which singles out a verb in the meaning of a construction becomes les dependable when it comes to ICV constructions. ICVs and their ICs tend to have a closer [verb +complement (s)] collocational and semantic association and thus seem to form a syntax-semantic unit in the lexicon than what obtains with [non-ICVs + complement(s)]. The following examples handle this situation:

In (12a and b), the verb kx?? has a non-ICV use and it has two meanings. In (12a), it selects two arguments and it is interpreted as shun meanwhile in (12b) it is understood as run away. However, in (12c), while in English, the verb is intransitive i.e having no complement, in Gh?maìlaì', there is a DP at the complement position. Moreover, by contrasting (12b) against (12c) one may notice that the interpretation of kx?Ì as «run» seems to totally depend on the following nominal element dù «race». This is the reason why Nwachukwu (1987), building his analysis on Igbo's data, argues that the verbal part does not contribute to the interpretation of the syntax-semantic unit by claiming that the verb root is meaningless. This semantic conception of ICVs and their ICs by Nwachukwu has an implication on his syntactic analysis of ICVs which will be discussed later in this chapter. Unlike Nwachukwu (1987), some scholars such as Essegbey (1999), Korsah (2011 and 2013) and Aboh (2015), adopting the well-known compositional semantics principle according to which, the meaning of a sentence is distributed among constituents of the clause, of which the verb is one, argue that ICVs have some consistent meaning which is traceable to the meaning of the entire clause. More precisely, using Essegbey (1999:252)'s terms, since the meanings of ICVs are underdetermined, their complement appear to further specify their meaning. Thus, it is clear that, for these scholars, both the verb root and the inherent complement contribute at some extent to the meaning of the complex although it is not obvious which part of the complex contributes most to the meaning. This rationale is also adopted in this work. This adoption is sustained by two empirical arguments/facts put forwarded by Essegbey (1999). First of all, if the verb root was meaningless and the meaning of the complex came from the ICs as suggested by Nwachukwu, we would expect ICVs having the same ICs to have same meaning. However, this is not often the case with ICVs having the same IC as shown below:

(13) a. Ba?ka^m k?Ì- w?ì s?ì nw?Ì
Bakam PST2-PROG _Vcountmatter «Bakam was thinking»

Data in (13) above clearly show that, although, the ICVs have a common complement nwà «matter» their meanings change depending on the verb root.

Another empirical fact that one may notice by observing the behavior of so-called regular ICVs, is that it is possible to capture the underlying meaning of ICVs which form a cluster. As I claimed earlier on the basis of data in (9) above, ICVs forming a cluster contribute at some extent to the meaning of the complex, each verb meaning is traceable to its non-ICV homophonous forms as illustrated in (7) and (8) above. So, considering what have been briefly explained above, it is plausible to assume that the verbs in Gh?maìlaì' ICV constructions have some meaning which is not as specific as non-ICVs in the language. Their meanings are generic and their complement appear to specify them as claimed by Essegbey. Accordingly, Aboh and Essegbey (2010:58) make a cross-linguistic observation according to which, languages that have ICVs also tend to have verbs that obligatorily take «semantically light» complements. This is the case for Gh?maìlaì' as shown in the following examples:

b. Ba?ka^m k?Ì-ts?ì ms?^ Bakam PST2-eat fufu corn «Bakam ate fufu corn» c.*Ba?ka^m k?Ì-ts?ì

Following Korsah (2011:106), ICVs cannot occur with generic-meaning complements like non-ICVs (see (14) above). Their meanings are derived either metaphorically or compositionally. Although it does not seem to be fully compositional (Aboh 2015), the meaning of some ICVs and their inherent complements can be traced to the literal meaning of at least one of the two syntax-semantic units. Korsah refers to these as Compositional ICVs. As for those whose meaning cannot be literally traced to any of the two constituents, he refers to them as Metaphorical ICVs. These are exemplified below:

In (15a) the meaning of the verb and the IC when put together may come from any of the two constituents. This is the case for nékx?Ì dù whose interpretation as «to run» mostly come from the meaning of the nominal constituent dù «race». The situation is slightly different in (15b) wherein the relationship between the verb and its IC does not suggest the meaning of the two together. For instance, néjô «to see» and ?wé «moon» put together does not suggest the meaning «to menstruate». Menstruations occur at the end of a cycle and announce the beginning of another cycle. Similarly, the moon appears at the end of a month and also announce the beginning of another month. This is the reason why, in Gh?maìlaì' and in most of the Bantu grassfield languages, speakers use the term moon to refer to menstruations. To menstruate metaphorically means to see moon in these languages.

It has been argued that verbs involve in ICV constructions are meaningful except that they are vague and less specific. Their inherent complements are required in order for them to be more specific in term of meaning. The inherent complement is thus the verb's meaning specifying part in an ICV construction. In Gh?maìlaì', most of the ICs are nominals, however, we can find a few number that are adpositionals (see the taìm cluster in (9) above). Accordingly, Aboh (2015) argues

that ICVs complements are structurally bare and therefore non-referential NPs. The inherent complement lacks a D-layer . In other words, the IC does not usually occur with determiners as shown in (16 and 17) below:

It can be noticed that when the ICs in (16b and17b) are construed as plural in an ICV construction, it results in ungrammaticality as compared to when they occur with a non-ICV as in (16c). I agree with Aboh that ICs lack the D-layer but will we observe that these NPs are referential in some context when I will address others morphosyntactic properties of ICs in the following section.

In some languages namely Igbo and Fon, the inherent complement has been described as a cognate complement. This means that the complement and the verb have the same form. In the aforementioned languages, some inherent complements are cognate with their verb roots as illustrated in the examples below:

This does not imply that only cognate ICs are attested; they are rather the exception. ICVs with non-cognate ICs are found in these languages (see Anyanwu (2012) and Essegbey (1999) for more details).

As for Gh?maìlaì', considering the examples cited so far (see data in (9), (10), (13), (15) and (16) above), there is no evidence that an ICV and its IC have the same form. However, the fact that Gh?maìlaì' is devoid of cognate complements does not necessarily mean that cognate object constructions are not attested in this language. The following example constitutes a proof that this type of construction can be found in Gh?maìlaì':

We can observe in (20a) that the verb fà2 and its complement fà2 are alike. The verb root can occur alone (without its complement); this is not the case for ICVs that obligatorily select a complement. After having described ICVs and ICs, the following section handles their morphosyntactic properties.

This section aims at bringing out morphosyntactic processes that either the verb or the complement can undergo. This is relevant not only in distinguishing ICVs from regular verb but also in discussing on whether the inherent complement is an argument of its verb or otherwise. In doing so, three morphosyntactic processes are addressed below namely pronominalization, focalization and question formation.

Pronominalization is syntactic process by which a noun is replaced by a pronoun. As a noun substitute, the pronoun carries the phi-features of the noun which it replaces. It is also the most evident and well known constituency test in the syntactic literature. An ambivalent

pronominal property with respect to the IC has been observed cross-linguistically. In some languages namely Ga, Akan (Korsah 2014), Gungbe (Aboh 2015) and Igbo (Anyanwu 2012), the inherent complement cannot be replaced by a pronoun. This operation leads to the production of ungrammatical structures as shown in the following examples from Igbo:

(20) a. EìzeÌ nÌtuÌruÌ maìiì

Eze pr.libate.past drink.IC «Eze poured a libation»

b. *EìzeÌ nÌtuÌruÌ yaì
Eze pr.libate.past it «Eze libated it»

(21) a. EìzeÌ nÌtuÌruÌ bóólu

Eze pr.throw.past ball.IC «Eze threw a ball»

b. EìzeÌ nÌtuÌruÌ yaì
Eze pr.throw.past it «Eze threw it»

(Anyanwu 2012:1563)

As Anyanwu (2012) reports, in Ngwa Igbo, the lexical NP of an affected object can be replaced by the pro-NP constituent, yaì «him/her/it', while the inherent complement cannot. This is the reason why the structure in (20b) is ungrammatical.

Data from Ewe in (22 and 23) below show that, in this language, the inherent complement can have a pronominal, just like nominal complement of non-ICVs. It is clear that Ewe is different from Igbo, Ga, Akan and Gungbe on this point.

(23) a.?? Nuìfiìaì laì n? anyiì haìfiì suku-viì-aì-wó n? anyiì.
teacher DEF sit ground before school-child-DEF-PL ICV ground.IC «The teacher sat down before the students did»

teacher DEF sit ground before school-child -DEF-PL sit-3SG «The teacher sat down before the student did»

Essegbey argues that the pronominalized form of the inherent complement is the preferred option when an ICV is repeated in subordinate clause such as in (23) above. Accordingly, he claims that the pronominalization of the IC is an indication that it is an argument of its verb.

As far as Gh?maìlaì' is concerned, the pronominalization of the inherent complement not seems to be evident to handle. When we observe the following data, we can deduce that the inherent complements cannot be replaced by the inanimate object pronoun já meanwhile the pronominalization of others complement is possible.

As shown in (24a-b) and (25a-b), when the inherent complement is substituted by the inanimate object pronoun já, the construction has a non-ICV interpretation. This suggests that the pronoun does not refer to the inherent complement. At this level, one can argue that IC cannot be pronominalized in Gh?maìlaì' and agree with Aboh (2015) that ICs are non-referential NPs.

Another fact about the language is that when a verb is repeated in a subordinate clause, the complement is either spelt out or deleted. This implies that the inherent complement cannot be pronominalized in this context like in Ewe. The following data are an illustration:

(26) a. Tâlaì k?Ì-pfâ bi?j?ì t? Ba?kâm pfâ
Tala PST2-eat groundnut before Bakam eat «Tala ate groundnut before Bakam did»

(27)a. Tâlaì k?Ì-t fwâ sî t? Ba?kâm t fwâ
Tala_PST2-Vsit ground before Bakam Vsit «Tala sat down before Bakam did»

The above data show that it is impossible to pronominalize the inherent complement when the verb is repeated within a subordinate clause (27b). Moreover, the deletion of the complement seems to be the preferred option when the verb is repeated in an embedded clause (26a and 27a). In addition, (26b) suggests that when the complement of the embedded clause is replaced by a pronoun, the latter refers to another entity which is not the complement of the main clause.

While observing two native speakers of Gh?maìlaì' interacting about the performance of an athlete, I gather the following data which suggest that the inherent complement can be pronominalized.

(28) Speaker A: ?a?làt mbaì?goÌ kxè dii! E l?-tâm dâ?zà n? ?ø ol?mpik
Charlotte Mbango Vrun race! 3SG.NOM PST3-come first on game Olympic «Charlotte Mbango runs! She was first during the Olympic games»

Já in the speaker B's utterance refers to dù «race». This pronoun carries both the phi-features and the semantic features of the inherent complement dù. If the IC is substituted by the third person singular animate object pronoun eì, the sentence will be interpreted as she avoids him. The data in (28) reveal that it is possible to pronominalize the inherent complement in Gh?maìlaì'. The pronoun is required in certain discourse contexts. The difference between the data in (25) and those in (28) is that the first ones have been collected through elicitation meanwhile those in (28) have been gathered within a natural setting. This is not a tentative to argue that data in (25) are wrong. As a native speaker of the language, I can claim that these constructions are attested in the language. In my opinion, the mixed pronominal property of ICs in Gh?maìlaì' can be justified by the fact these complements tend to lose their nominal properties. This is the reason why it fails to be pronominalized in (25). Although the IC can be pronominalized in certain contexts, we cannot straightforwardly argue that it is an argument of its verb. Indeed, given the fact that there is not a one-to-one mapping between case features and semantic roles, theta role assignment cannot been tied to case features. Expletive constructions in English strengthen this observation. Accordingly, the structural position of subject filled by the expletive it in those constructions enables this latter to bear nominative case. However, it is a shared knowledge that the expletive it bears no semantic role.

Previous studies on Gh?maìlaì' (Tala 2015) reveal that both arguments and adjuncts can be focalized in this language. Two focus positions have been identified in the language: the C-domain and the v-domain. Left-peripheral focus encodes new information whereas post-verbal focus deals with contrastive focus. The examples below are instances of focalization.

As shown in the above data, left-peripheral focus is marked by two particles namely né and té which structurally follow the focused constituent. Né is used for subject meanwhile té is used for non-subject focus. Data in (29b) and (29c) are instances of information focus whereas those in (29d) and (29e) are cases of contrastive focus. Low focus is either marked by pé or aì that precede the focused item. The following table presents in a sketchy fashion non-verbal constituent focalization in Gh?maìlaì':

Focus markers	Focus Strategies	Domains	Fonctions	Constituents
né /lé	Ex-situ	C-domain	Informative	Subject
t?ì (with clefts)	Ex-situ	C-domain	Informative	Non-subject (objects and adjuncts)
aì/p?ì	«In-situ»⁸	v-domain	contrastive	Non-subject (objects and adjuncts)

As far as verb focalization is concerned, Gh?maìlaì' verb focus constructions have a null operator and involve predicate doubling structures. Unlike Gungbe (Aboh and Dyakonova 2009) and Ga (Korsah 2014) wherein the focused copy is fronted sentence-initially and the other copy is found in the extraction site within the IP; the two copies occur IP-internally in Gh?maìlaì' as shown in the examples below:

As for the focalization of the verbs in constructions involving ICVs, both the verb and the inherent complement can be focused in Gh?maìlaì'. Just like non-ICVs, inherent complement verb focus constructions involve verb doubling structures in which the two copies of the verb appear within the I-domain. These structures have a contrastive reading. Let us consider the data in (31) below:

«Bakam BREAST-FEEDS the baby before going to church» (as opposed to say, she rocks the baby)

Data in (31) show that ICVs can be focalized in Gh?maìlaì' just like in Basaa (Bassong 2014: 286). This focus property of ICVs distinguishes Gh?maìlaì' from Kwa languages. Indeed, studies of ICVs constructions in these languages reveal that inherent complement verbs can neither be fronted nor doubled in verb focus as shown in the following data:

«Kwei ran (as opposed to say, he sat)» (Ga, Korsah 2014:410)
It can be observed in (32) that the focalization of ICVs is infelicitous in these languages.

As it has been claimed earlier, the inherent complement can also be focused in Gh?maìlaì'. The IC is thus preceded by the focus marker p?ì or aì. Let us observe the data below:

When we compare the data in (33) with the one in (29d), it can be noticed that the focused element is the same in both constructions i.e., the object/nominal complement. What fundamentally distinguishes constructions in (33) from the one in (29d) is the semantic interpretation. Indeed, the interpretation conveyed in (33b) is that, apart from spiting, Tala does not know to do something else or has not done something else. We can therefore deduce that in constructions such as those in (33) above where the focused item is the IC, syntactic focusing does not lead to their semantic focusing. More precisely, the focusing of the IC results in a predicate focus. A focused IC has a predicate focus reading, not an argument focus reading. A predicate focus interpretation of the focused IC seems to be a cross-linguistic feature of ICV constructions. Data from Gungbe and Ga show that these languages exhibit this IC focus property as shown in (34) below:

As one can observed in the above data, the focusing of the IC results in a predicate focus. This suggests the existence of a closer relationship between the IC and its verb. The fact that ICVs can be focused in Gh?maìlaì' may be an indication that ICV constructions are not so different from regular verb construction. As for focalization, the sole distinction between ICV verbs and non-ICV ones is that object focus has argument focus reading in non-ICV construction whereas it has a predicate focus reading in ICV one.

According to Tala (2015), three types of questions are attested in Gh?maìlaì' namely yes/no question, tag question and wh-question. As for wh-question, the wh-expressions identified in the language comprise wá «who», k? «what», sóó «when», hé «where», m gá^k?Ì «how» and n?k?ì «why». These wh-operators can either be extracted or left in-situ. When they move instead, they are fronted and therefore generate automatically a focus construction or a relative clause. Wh-question is more adequate here since it lays emphasis on constituents of the clause which will be the inherent complement in this work.

Question feature has been acknowledged by Korsah (2014) as a morphosyntactic feature that sets ICVs apart from lexical verbs. The inherent complement cannot be marked with a question feature. Let us consider the following data:

(35) a. Ta^laì k?Ì-pfâ bi?j?ì
Tala PST2-eat groundnut «Tala ate groundnut»

Unlike the complement of non-ICVs (35), the inherent complement can neither be marked with a

question feature in-situ nor be extracted to the left periphery for focus wh-question formation.

(36) a. Ta?laì k?Ì w?ì kxù di?
Tala PST2 PROG _Vrunrace «Tala was running»

The data in (36) show that when the inherent complement dù is substituted by the argument wh-expression k?, the construction itself is legible but what fundamentally changes is its semantic interpretation. Indeed, the structures in (36b and c) are infelicitous if and only if the predicate is interpreted as «to run». When the speaker utters sentence (36b), even if the agent is running meanwhile the conversation holds, the verb root kxù is no more understood as «to run»; it is rather interpreted as «to run away» or «to avoid». So, the ICV meaning «to run» is lost in (36b). However, (37a) can be an appropriate answer to an interrogative expression with a more generic-meaning such as (37) below:

This is an indication that ICV verbs are semantically light (Korsah 2014), vague or less specified as reported by Essegbey (1999). Moreover, as claimed by Korsah (2014), the inability of IC to be marked with a question feature may be an indication that it is not an argument of its verb assuming that the question feature is marked on complements of the verb which are arguments.

In the previous section, it has been shown that the IC can be realized as an overt pronoun in certain discourse context (see (29) above). It has also been demonstrated that both the ICV and its complement can be focused. However, when the IC is focused, it has a predicate focus reading. Moreover, the IC cannot be marked with a question feature, be in-situ or extracted. These morphosyntactic properties that distinguish ICVs from non-ICVs have an impact on the argument structure and the derivation of ICV construction adopted here.

Earliest proposal on the argument structure of ICVs was done by Nwachukwu (1987) based on data from Igbo. He analyzed the inherent complement as an adjunct. His rationale was motivated by the fact that it is easily displaced when the ICV selects an internal argument as shown in (38) below and this displacement of the inherent complement explains why it does not bear a semantic role just like adjuncts.

In (38b), the internally licensed argument is OÌbiì. According to Nwachukwu, in (39b), the IC has been displaced and he treated it as Move IC. It moves rightward as illustrated in the following tree structure:

Nwachukwu's rationale need to be revisited for two reasons: (i) rightward movement is prohibited by modern approach to syntax, more precisely, by the antisymmetric approach (Kayne 1995); (ii) some adjuncts might have semantic role as it is the case with English passive construction wherein the agent is introduced with a by-phrase. However, Nwachukwu's work has the merit of having raise an essential point in any discussion on argument structure of ICV namely, the status of the

inherent complement. To be more explicit, we want to know whether the inherent complement is a semantic argument of its verb or otherwise.

On this point, two camps of scholars have emerged in the literature. The first one is made up of those who claim that the IC is an argument of its verbs just like any complement of non-ICVs. This view is strongly defended by Essegbey (1999, 2002, 2003 and 2010) who claims that ICs are arguments of ICV and the sole difference between them and complement of non-ICV is that they are obligatory complement; they need to be cited for the unit to be meaningful. So, its obligatoriness is not from syntactic relevance but from semantic relevance. His opinion relies on the fact that IC can be realized as an overt pronoun in Ewe and therefore it bears both case features and phi-features (see (22) above). As it has been said earlier, there is not a one-to-one mapping between case features and semantic roles, theta role assignment cannot been tied to case features. The second camp of scholars gather together those who argue that the IC is a syntactic argument not a semantic argument of its verb since it does not bear a theta role (see Korsah (2014) and Aboh (2015)).

As for Gh?maìlaì', given the morphosyntactic properties that the IC exhibits in this language, it can be infered that the IC is a syntactic argument not a semantic argument of its verb. Two facts support this rationale:

On the basis of these facts, it is evident that the IC cannot bear a theta-role since this latter is assigned to argument ( see Chomsky (1981)). It does not allow certain syntactic processes that are typical of arguments. The VICV thus does not have an internal argument. It tends to be functional and shares a common semantic characteristics with light verbs.

Assuming Pesetsky's notions of C(ategorial)-selection and S(emantic)-selection, Korsah (2014) argues that there is a two-level lexical entry for every verb ( whether it is a full lexical verb or an ICV): one level deals with syntax (Syn = C-selection) and the other deals with the semantics ( Sem = S-selection). There is full match up between the syntax of a construction and its semantics when Syn and Sem are both accessible to the verb and its complement. Accordingly, he claims that what typically happens in ICV constructions is that, there is only a partial match between the

verb and Sem. The verb is syntactically represented and morphologically spelled-out but it lacks the needed semantics. This is the reason why the meaning of what the predicate denotes is mostly closest to the IC and the ICV cannot assign theta role to its nominal complement since theta roles are assigned to semantic arguments which would be found in Sem. This configuration explains why the IC though might show case and phi-features, is not an argument of its verbs in the same sense as the argument of lexical verbs. The distinction between ICVs and full lexical verbs (FLV) in terms of Syn and Sem properties has been graphically represented by Korsah (2013: 417) as follows:

As reported by Korsah, it would be problematic to claim that the verb is totally delinked from Sem since not all the meaning of the predicate in an ICV construction might be from the complement. As it has be shown above, the verb root contributes at some extent to the meaning of the predicate (see § 4.1.1.2). This partial link between ICV and Sem is represented in (40ii) by a dotted line. The dashed line linking the ICV and the IC in (40ii) is a way for him to indicate the verb phrase reflexes on the IC in certain syntactic operations such as focus construction wherein the focalized ICs have a predicate focus reading. Given the above discussion, it is evident that most of the ICVs in Gh?maìlaì' are intransitive since most of them select only one argument which is the external argument; few are transitive.

Each element of the [V-N] complex comes from the lexicon free. The fact that a focus marker can occur between the IC and the ICV (see (33) above) is an evidence that the verb and its complement are not bound in the lexicon. In other words, the ICV and its IC enter into the derivation as two different syntactic units just like any other verb root though semantically bound. Following (Aboh 2015), it can be argued that there are two types of verb roots in the lexicon: those that can directly merge into a functional domain namely under ?° and the ones which merge under

V° and latter move to ?°. ICVs just like light verbs are functional verbs and therefore they can purely merge under ?°, the head of light verb phrase (?P).

The occurrence of a focus marker between the verbs and its complement rules out the derivation proposed by Essegbey (2010) in which ICV is merged under V° and it takes a bare NP which latter incorporates to V° wherein the verb and its complement form a compact unit and latter move to v°. Essegbey's proposal also prevents the possibility for an element of the [V-N] complex to be a target of a syntactic operation. As it has been shown above (see section 4.2.2), both the ICV and its complement can be subjected to focus operation as two independent syntactic units in Gh?maìlaì'.

The linear adjacency between the ICV and its complement (see (1 and 2)) in Gh?maìlaì', as opposed to Igbo wherein an argument can appear between the verb and its IC (see (38b)), favours an analysis in which the V selects a structurally bare NP. Moreover, the incorporation of the IC to V is favored by the fact that focused IC has a predicate focus reading. Let us observe the tree diagram in (41b) below which the representation of the construction in (41a).

Following Korsah (2014) and Aboh (2015), merging the verb under t° as opposed to V has a number of conceptual and empirical advantages. In fact, by generating ICV under t° it has no s-selectional requirements on the complement and it introduces the external argument. Although the abstract V c-selects a structurally bare NP as complement, it does not have any theta-role to assign. With this analysis, any element of [V-N] complex can be the target of a syntactic operation as shown in the examples below:

(43a) is the tree representation of sentence (42a). In (42a), Ta^la?, being the external argument is licensed by t and therefore is merged under the specifier position of the light verb phrase. It latter moves to Spec-TP in order to satisfy the EPP requirement of TP. The Aspect Phrase projected above the small VP represents the aktionsart of the verb kxù. Being a lexical property, the right

position of this phrase is immediately above the big VP. The actual position in the diagram is due

to the fact that the verb itself is directly merged under ?°. Given the copy theory of movement, it would be problematic to argue that Gh?maìlaì's verb doubling involves two spelt-out copies without

theoretical evidence. So, following Aboh and Dyakonova (2009)'s parallel chain analysis of verb doubling, I argue that verb movement in Gh?maìlaì'verb doubling construction is triggered by two

probes namely Foc° and Asp°, the head of the Aspect Phrase which represents the aktionsart of the verb. The two probes have the same goal. This is the reason why Gh?maìlaì' verb focus involves

The tree diagram above represents the construction in (42b). In (42b), the sentence has a predicate focus reading though syntactically it is a nominal element that is focused. This is an indication that it is not the bare NP complement which is focused but the entire abstract VP with the incorporated NP to V that undergoes focalization as represented in (43b). There is an asymmetry in Gh?maìlaì' focus strategies. In the left peripheral strategy, the focused item precedes the focus marker meanwhile in the v-domain strategy the focused constituent follows the focus marker. Bearing in mind fundamental basics of Rizzi (1997)'s rationale according to which, the focused constituent

should be hosted by the specifier position of FocP, I generate the so called focus marker under the specifier position of CleftP which dominates the FocP. By generating the focus marker under Spec-CleftP, the verb can freely merge from ?° to Asp° passing through Foc° and Cleft°. The verb movement is followed by the pied-piping of ?P to the specifier position of FocP, the position dedicated to focalized constituent, and the right order of constituents is derived.

The morphosyntax of ghomala' verbs: focus on inherent complément verbs and serial verb construction

Conclusion

Introduction