Difference between revisions of "Multi-verb constructions in Edo"

Revision as of 18:27, 15 June 2011

By Ota Ogie Researcher affiliated with the Department of Language and Communication Studies. NTNU E-mail:ota.ogie@hf.ntnu.no

Introduction

This study examines multi-verb constructions in Èdó (a Benue-Congo language) with the aim of identifying and classifying them and their argument sharing patterns. Èdó is spoken in Èdó state in Mid-Western Nigeria and belongs to the Edoid language group (Elugbe1979). It is a head initial SVO language with an open syllable system with no consonant clusters.

We draw main background assumptions from the following sources; implemented Head-Driven Phrase Structure Grammars for Norwegian (Hellan 2003) and Ga (Hellan 2007) a Kwa language spoken in Ghana; an HPSG account of argument realization patterns (Beermann, Hellan and Sætherø 2003); and a theory of event structure template (Pustejovsky 1995, 2006).

Multi-verb constructions provide useful insight into the question of how languages distinguish between adjunction and complementation. The term multi-verb constructions is defined as consisting of verbs in series that can function as independent verbs in simple sentences, with at least one shared argument and no marking of syntactic dependency (cf. Amaka 2005:2):

1. Èdó

sv-v1objIDv2su-v1tr-v1suAg-v1obAff-v2-int-v2suAff-CAUSE_RESULT

The template above the construction (Hellan and Dakubu 2009, Ogie 2010) gives the construction labels and has the following structure: Area 1 gives the global labels, the number of verbs in series (ie sv, sv3, sv4 ) as well as argument sharing information and information about thematic relations holding across the verb in series. In the example above sv states that the construction is of type serial verb with 2 verbs, Area 2 gives the valence information as well as information about grammatical function and thematic roles. This is exemplified by v1objIDv2su-v1tr-v1suAg_v1obAff-v2-int-v2obAff in the example above. Here the object of v1 shares reference (ID) with the subject of v2 v1objIDv2su. Also v1 is transitive and its subject has an agent thematic role and the object an Affected role v1tr-v1suAg-v1obAff while v2 is intransitive with a subject having an Affected role v2-int-v2suAff. Information about the situation type of the construction is provided by Area 3 and this is exemplified by CAUSE_RESULT in the above example. Area 3 is written in capital letters.

In examples where there is token identity between arguments of the verbs in series, I have represented thematic relations for such examples only in area 1. Area 2 then gives information about valence, arguments that do not share thematic roles and arguments that are not shared. For example, in the consequential svc in the Èdó example below, the template reads as follows: Area 1:sv_suObID_suAg. The serial verb construction consists of two verbs in series and the verbs in series share reference across arguments. The subject argument is token identified and has an agent theta role. Area 2:v1tr-v1obThincrem-v2tr-v2obAff. V1 is transitive and has an incremental theme object. V2 is also transitive with an object bearing an affected theta role. Area 3: The situation type is a TRANSFER relation.

Information about Tense Aspect, Mood and Negation is also provided by area 1 in the construction labels. Sharing of these features across verbs in series is represented as with sharing of arguments. For example in the Akan example (example 10) on covert reference subject sharing in the clause chaining serial construction below , area 1 gives the following information sv_suAspID_suAg_aspCompl. It states that the construction is a serial verb construction with the verbs in series sharing subject and aspect values. The subject has an agent thematic role and the verbs in series have completive aspect.

With respect to the global labels in area 1, Hellan and Dakubu 2009 uses the global label ev to represent Extended Verb Complexes and the label pv for preverbs in EVCs. In addition, to the labelling conventions used by Hellan and Dakubu 2009 for SVCs (sv) and EVCs (ev, pv) , the following global labels are introduced to account for the range of multi-verb constructions in my data. The background assumptions remains the same (see Hellan 2008 and Hellan and Dakubu 2009 and In-depth annotation of multi-verb constructions in Èdó for discussion on the labelling conventions:

Global labels

consecutive construction - csc

covert coordination - cc

infinitival construction - ic

modifier construction - mc

negative resultative construction - nrc

overlapping construction - oc

complement/embedded clause - cec

empty subject construction - esc

Verbal Morphology

With respect to verbal morphology, verbs can be inflected for the purpose of pluralizing nouns they occur with and to mark iteration.

Simple sentence: V+plural suffix

v-tr-suNom_ag_obThincrem-COMPLETED_MONODEVMT

2.

Tense and transitivity are marked on the verb either through tonal changes or by affixation of a past tense suffix –rV under appropriate licensing conditions. With plural verbs,the order is the plural suffix before the past suffix.

4. Simple sentence: Present tense

v-intrImplobj-suNom_ag-NONCOMPLETED_MONODEVMT

5. Simple sentence: Past tense

v-intrImplobj-suNom_ag-COMPLETED_MONODEVMT

Multi-verb constructions:Tense,Aspect,Mood, argumentsharing and situationtype

The verbs in series in multi-verb constructions need not bear one/same marking for tense, aspect, mood or negation and need not share subjects. They include SVCs, consecutive constructions, covert co-ordination, overlapping constructions and V+ infinitival complement constructions.

Èdó

6. Consequential SVC

sv_suObID_suAg-v1tr-v1obThincrem-v2tr-v2obAff-TRANSFER

7. Covert co-ordination cc_suID_suAg-v1tr-v1obAff-v2tr-v2obThincrem-CAUSE_RESULT

8. V+infinitival complement construction

ic_suID_suAg-v2tr-v2obThincrem-CAUSE_RESULT

9. Ewe

Consecutive constructions

cs3-v1intr-v1suAg-v2intr-v3intr-v3suAg-DIRECTED_MOTION

Clause chaining serial constructions

Akan

10. sv_suAspID_suAg_aspCompl-v1tr-v1obAff-v2intr-CAUSE_RESULT

11. sv_AspID_aspCompl-v1ObIDv2su-v1tr-v1suAg-v1obAff-v2tr-v2suAff-CAUSE_RESULT

One of the verbs in sereis in V+modifier constructions is shown to be reanalyzed as an adjunct. The construction has one marking for tense, aspect and mood.

12. V+modifier constuctions

mc-vintr_suAg-ACHVMNT-MOTION_DIRECTED

11 verb constructions in Èdó are shown to pattern into four structural types with respect to the distribution of the past tense suffix –rV, an infinitival marker yá, a floating anaphor tòbórè 'by him/her/it self ', VP adverbs and argument sharing patterns. Of the 11 verbal constructions 7 are shwon to be multi-verb constructions:

A. V (P) +V (P) constructions: resultatives, negative resultatives, consequential and covert co-ordination constructions: -rV not licensed, infinitival yá not licensed. The verbs in series have the same values for Tense, Aspect and Mood (TAM).

B. V + mood constructions: purpose constructions: -rV licensed, infinitival yá not licensed. V2 has a positive value for MOOD.

C. V+ infinitival complement constructions: comitative and instrumental constructions: -rV licensed, infinitival yá licensed and V2 is non-finite.

4 of the 11 verbal predicates are shown to be reanalyzed V+ modifier constructions: D. V+ modifier constructions: durational, directional, locational, manner constructions: -rV licensed, infinitival yá not licensed. One verb in the series is reanalyzed as adverb.

The –rV suffix also interacts in an interesting way with the temporal structures of multi-verb constructions. Overlapping events license –rV while non-overlapping events do not

A typology of multi-verb constructions

The study examines multi-verb constructions in the following languages of the Niger-Congo: Igbo and Yoruba (Benue-Congo), Gurenne (Oti-Volta), Ga, Baule, Akan and Ewe (Kwa) and situate properties of Èdó multi-verb constructions within typology common to these languages.

ÈDÓ

Infinitival complement construction

ic_suID_suAg-v2tr-v2obThincrem-CAUSE_RESULT

Consequential SVC

sv_suObID_suAg-v1tr-v1obAff-v2intr-v2obThincrem-CAUSE_RESULT

IGBO

Commutative SVC

sv_suID_suAg-v1tr-v1obTh-v2intr-DIRECTED_MOTION

AKAN

Clause chaining SVC (CCSVC)

sv-Implobj_suObAspID_suAg-v1tr-v1obAff-v2tr-v2obThincrem-CAUSE_RESULT

YORUBA

Commutative SVC

sv_suID_suAg-v1tr-v1obTh-v2intr-DIRECTED_MOTION

EWE

Complement/embedded clause construction

cec-v1intr-v1suAg-v2tr-v2suAff-v2obThincrem-CAUSE_RESULT

BAULE

Empty Subject Constraint (ESC)

esc_suID_suAg-v1tr-v1obThincrem-v2tr-v2obAff-CAUSE_RESULT

GURENNE

Theme SVC

sv-suObID_suAg_obTh-v1tr-v2tr- PLACEMENT

Ga

Extended Verb Construction

ev_ditr_suObID_suAg_obTh-pv1tr-vditr-iobEndpossessor-TRANSFER

Multi-verbs identified include SVCs in all the languages discussed, consecutive constructions and overlapping constructions in Ewe and covert co-ordination in Èdó, Igbo and Baule. Typological features used for identification include: tense, mood, aspect, negation, adverb distribution, predicate cleft and argument sharing patterns. The findings show that the typological features of a language determine the type of multi-verb construction it licenses. Also while, inflection may demarcate multi-verb types within a language, the pattern observed for a language may not map onto another language.

With respect to argument sharing, the following types are discussed; token sharing of subjects; switch sharing; overt reference sharing of subjects; covert reference sharing of subjects; token sharing of objects ; overt reference sharing of objects and covert sharing of objects. The patterns found in the languages studied support the claim that languages with rich verbal agreement features allow recoverability of unexpressed arguments and tend to license null subjects and objects. Object sharing patterns show asymmetry with respect to switch sharing and reference sharing. Languages that have overt reference subject sharing patterns do not have switch sharing (Ewe, Ga and Baule) while those that do not, tend to employ token/covert reference sharing of subjects and switch sharing (Èdó, Yoruba and Akan).This is buttressed by data from Attie and Likpe closely related languages to these languages. With respect to object sharing, these languages that do not have switch sharing all have covert sharing of objects, while those that have, do not have covert sharing of objects. Èdó belongs to the type that does not have overt reference sharing of subjects and tend to employ token sharing of subjects and switch sharing. For object sharing, Èdó does not have covert sharing of objects and employs mainly token sharing of objects.In particular, object sharing in multi-verb constructions in Èdó is analyzed as token sharing by grammatical function.

Schema for multi-verb constructions

Two schemas are posited to account for Èdó multi-verb constructions:

1.Verb-serial-compl (ement)-phrase with a complementation structure for the V (P) +V (P) resultative and V+infinitival complement constructions.

2.Serial-mod-phrase with an adjunction structure for V+mood constructions, V+modifier constructions and V (P) +V (P); consequential, purpose, and negative resultative constructions.

An Èdó GrammarMatrix

Ota Ogie is developing an Èdó GrammarMatrix based on her Ph.D dissertation Multi-verb constructions in Edo (Ogie 2009a).The grammar is constructed on the LinGo GrammarMatrix (Bender, Flickinger and Oepen 2002, Bender et al 2010) and is based on the Head Driven Phrase Structure Grammar (HPSG) (Pollard and Sag 1994, Sag and Wasow 1999, Sag, Wasow and Bender 2003) and the Linguistic Knowledge Builder (LKB) system (Copestake 2002, Copestake et al 2005). Assumptions are drawn from the NorSource GrammarMatrix (Hellan 2003, Hellan and Haugereid 2003, Hellan and Beermann 2006 etc) and the Ga GrammarMatrix (Kropp Dakubu M.E., Lars, Hellan, and D.Beermann. 2007, Hellan 2007). The grammar performs both parsing and generation and accounts for phenomenon like basic clause syntax, agreement, tone, Tense Aspect and Mood, modification and MVCs(Ogie 2011)media:Edo GrammarMatix.zip.

References