Ga Extended Verb Complex

Grammatical Feature Sequencing in the Ga Extended Verb Complex: a Formal Approach

Dorothee Beermann Lars Hellan and Mary Esther Kropp Dakubu

In Beermann, Hellan and Dakubu (2004) we examined the Ga extended verb, referred to in that work as the "Preverb Complex" but which we now prefer to call the "Extended Verb Complex" (EVC), on the grounds that the new term better describes the construction as a whole. (Any resemblance to the term SVC is not accidental.) A pre-verb pattern can consist of 3 different types of pre-verbs, leading to the following schema (BH&D 2004: 105 fig.16). Only the rightmost ("main") verb is obligatory.

(BH&D 2004: 105 fig.16).

In BH&D (2004), using Head-driven Phrase Structure Grammar as our formal framework, we:

• set up a type hierarchy of verbs entering into the construction;
• outlined the form of a sign for the preverb;
• argued that valency propagation requires recognizing the leftmost (pre-)verb, and not the lexical (rightmost) main verb (unless it occurs with no pre-verbs), as

the head of the construction;

• elaborated the signs for two types of pre-verb (Vk, Vdeictic).

In conclusion, we claimed that these constructions should be "…analyzed as recursive head complement structures, constituting a single word, albeit with dependent word forms as constituents. The boundedness of the pre-verbs to the PreVC construction is analytically expressed through the obligatoriness of their complement V …. The head-complement rule is in essence the same as what one expects at phrasal level, and this hybrid nature of phrasal syntax and semantics and word-internal morphology and phonology is seen as capturing the intermediate status of pre-verbs as phenomena situated between syntax and morphology."

In this paper we pay closer attention to an important feature of the construction, namely that the Aspect-Modality-Polarity features of the EVC as a whole are marked in a fairly intricate way, such that the EVC as a whole can be said to have a single value for each of Aspect or Modality and Polarity, but depending on which pre-verbs are present the value may

• be marked on a different constituent, and
• there are differing constraints on possible values.

For this squib we will focus on the fact 2 mutually exclusive features, namely ASP and MOD, and we contend that a declarative EVC is marked for one and only one, or none, of three Aspects: DYNamic, COMPLetive, and PROTRacted.In some circumstances the feature PROTR may be realized by one of two features, PROGRessive and HABitual, but in others only HAB is available. These features correspond to the highlighted nodes of the tree display of the inflectional category types, reproduced from BH&D (2004: 99). The feature COMPL corresponds to the node labelled perfect, PROTR to the node labelled non-perfect, and DYN to the node labelled volitional.

(2) Ga Inflectional category types :

There are however important differences between the feature hierarchy as diagrammed and the hierarchy deployed in this paper. The design of the tree and particularly the positioning of the volitional node under the irrealis node were motivated in large part by sequencing phenomena of the SVC. For the declarative EVC, on the other hand, all three nodes are more efficiently viewed as dominated by ASPect. The other features relevant to the EVC are dominated by NON-VOLitionality and are conveniently dealt with separately. For ASPect therefore the heirarchy is almost flat: (3) Ga aspectual category types:

Insert IMAGE

A major advantage of this schema is that it represent the Negative Declarative EVC as well. In the negative, each of COMPL, PROTR and DYN has a single and distinct phonological representation (PROGR and HAB as well as EVCs with no aspectual marking, ie. all nodes dominated by non-perfect in (2), have identical forms in the negative.

Although the PROTR / non-perfect node dominates progressive and non-progressive, most saliently habitual, on the tree, we will show that, except when the only pre-verb is k, which as we shall show is exceptional in other ways, in an EVC, PROTR can be treated as coincident with habitual.

In (2), non-perfect│non-progressive also dominates the feature aorist. This feature, which it may be noted has negative values all the way up the tree, is now treated not as a feature but as a default form in the absence of one of the three declarative features. A verb not marked for any of these features is interpreted as past or present, depending on the verb and the circumstances of the speech event.

We now state the general rules for feature marking. For all declarative EVCs:

• not more than one of the three features is phonologically marked (presence of one excludes the other), ie. only one can have a positive value.
• only one verb in the complex takes a feature marker.
• Vk is never phonologically marked for a positive value for any feature.
• The (positive) markers of the following features occur on the right-most verb of the EVC: DYN; PROTR; all 3 features in the presence of NEG.2
• One feature is marked on the penultimate (left of rightmost) verb: COMPL (non-NEG) (In view of these and the constraints listed below, this formulation is redundant – since

k is never marked and ka doesn't permit ASPect, COMPL can be said to be on the leftmost available verb.)

Non-declarative ie. NON-VOLitional EVCs are rather different. The relevant features descend from the non-volitional branch of the tree (1). Apart from adding a suffix, the plural-imperative is exactly like the subjunctive, and so is the non-root-imperative except that it has no subject marker. The salient feature is thus that labelled subjunctive, which we shall call SUBJunctive.

Unlike the declarative features, if there are at least two verbs present NON-VOL is marked twice, on the two leftmost verbs other than k. Furthermore, one EVC can be marked by both of the features displayed below. Whichever feature is selected for the first of the verbs to be marked, the second one always carries SUBJ.

(4) Features of the Non-Volitional EVC:

NON-VOL ┌──┴───┐ SUBJ IMPER subjunctive direct imperative

The plur-imper marker (-a) is treated as a contextual variant of the direct / root- (singular) imperative, because both are explicitly direct IMPERative markers, are suffixes, and presence of one excludes the other. Figure (2) can now be replaced by (5) (5) Ga inflectional categories: MODALITY ┌─────┴─────┐ ASP NON-VOL ┌───┼────┐ ┌──┴──┐ COMPL PROTR DYN SUBJ IMPER ┌──┴──┐ PROGR HAB

We can now state feature sequencing constraints on EVCs. It turns out that constraints are simply stated as progressively introduced by the pre-verbs, moving from left to right: 1. If Vk is selected, no constraints are introduced. At the same time, on this particular verb all features have negative value, a local constraint which is not carried on to the next verb. 2. If VNEG is selected (whether or not preceded by k, and no matter what follows) only NON-VOL is available. 3. If Vdeict is selected either NON-VOL or ASP may be selected. All features are available except PROTR │PROGR – bearing in mind the placement rules for declarative EVCs above. 4. If no preverb is introduced, no constraints apply. Point 1. allows expressions in which a phonologically heavy prefix occurs between k and the main verb, eg. (6) e-kɛ m-ba he is bringing it 3S-k PROGR-come which is not allowe after Vdeict. Point 2. implies that cases (6) cannot thereafter arise, as does 3.

PROBLEMS: IMPER is a suffix and expressed only on the rightmost verb, except that with ka, in the neg. direct singular imperative, it is suffixed to ka (and the next verb is SUBJ according to rule), ie it is very much a mini-SVC. Note that if there is just one preverb other than k, the rightmost (main) verb is marked in the plural imperative by both of the NON-VOL features. Another good reason to separate ASP from NON-VOL. There appears to be a real break between k and the other verbs – not only can an object come between them, a non-contracting prefix can follow on the next verb (which is necessarily the rightmost verb - the two sets of constraints take care of this I think). Does the right branching tree diagram (13) in the last paper take care of this adequately? The fact that k and the main verb in themselves introduce no constraints on feature selection, but ka and to a lesser extent the deictics do, suggests to me that at some level k and the main verb are equivalent and the other two go under the main verb – I guess notionally? In any case, the thesis that (13) is syntactically correct and not (16) seems vindicated. Also, a primary distinction between k and the rest of the EVC tallies with what happens in non-initial and Vbid EVCs.

References Beermann, Dorothee, Lars Hellan and M.E. Kropp Dakubu 2004. The Ga 'preverbs' k, ya and ba from an HPSG perspective. In Dakubu and Osam, eds., Studies in the Languages of the Volta Basin 2: 94-112. Legon: Linguistics Dept.

Media:Grammatical Feature Sequencing in the Ga Extended Verb Complex.doc This paper was prepared for a Legon-Trondheim Linguistics Project colloquium held at the University of Ghana. It is not exactly the publication version, but very close.