Simple and complex verbs in Jaminjung. A study of event categorisation in an Australian language.

This thesis is an investigation of the semantic and syntactic properties of simple and complex verbs in Jaminjung and Ngaliwurru, two closely related varieties of an Australian language (henceforth simply ‘Jaminjung’). This language exhibits a typologically unusual characteristic: it has two distinct parts of speech in the function of verbs. One of them, termed ‘(generic) verbs’ here, is a closed class with around 30 members, which obligatorily inflect for person and tense/aspect/mood. These may constitute predicates on their own in a finite clause. ‘Coverbs’, on the other hand, constitute an open class; they are inherently predicative, but do not inflect. Coverbs may function as the sole predicate only in non-finite dependent clauses. In finite clauses, they only occur as part of a complex verb, together with a generic verb. An interesting consequence of this ‘division of labour’ between coverbs and generic verbs is that the verbs categorise events. Since a verb from the closed class is obligatory in every finite clause (either as a simple verb or as part of a complex verb), all event expressions are sorted into one of a small number of categories by the choice of generic verb. This type of overt categorisation is comparable to that found in the domain of entity expressions in languages with nominal classification. Two major, interrelated questions are addressed in this study. The first question concerns the conceptual basis for the categorisation of events by generic verbs. The second question concerns the syntactic and semantic conditions on complex verb formation, or, put differently, the contribution of verb and coverb, respectively, to the interpretation of a complex verb. Chapter 1 contains background information on the language and its situation. Jaminjung is spoken today only by around 100, mainly older, people, and no longer acquired by children. Written documentation is very scarce, and this thesis is based mainly on the fieldwork of the author. The theoretical and methodological basis for the investigation are also laid out in the first chapter. It assumes a construction-based model of grammar, that is, grammatical structures are regarded as schematic complex signs. For the semantic analysis, monosemy is assumed as a heuristic principle. Polysemy is recognised where necessary, but where it is possible, more specific interpretations are derived from contextual information or general pragmatic principles. Chapter 2 describes the essential grammatical features of Jaminjung insofar as they are necessary for the understanding of the examples and the argumentation. In this chapter it is also demonstrated that coverbs constitute a major lexical category whose members differ not only from generic verbs, but also from nominals in their morpho-syntactic properties. Chapter 3 deals with simple and complex predicate constructions. Simple verbs consist of a generic verb functioning as a predicate on its own. A comparison of the text frequency of the different constructions reveals that in 40% of all finite clauses the predicate is a simple verb, despite the small number of verbs that are available in this function. Canonical complex verbs are defined as the combination of one or sometimes two unmarked coverb(s) and a verb, in a single intonation unit. It is argued that complex verbs are both part of the lexicon (in that they are, in their majority at least, conventionalised expressions), and part of the grammar (in that they are semantically compositional and instantiations of a productive construction). Closely related to the canonical complex verb construction is the progressive construction, where the coverb is marked as ‘continuous’. Canonical complex verbs are also distinguished from constructions where the (marked) coverb functions as a secondary predicate, or where an (unmarked) coverb functions as a ‘semi-independent’ predicate in an intonation unit on its own. In a final subchapter, strategies of integrating loans from Kriol – an English-based creole spoken in Northern Australia – are discussed. A frequent strategy is to employ Kriol verbs as coverbs; these forms, just like Jaminjung coverbs, are combined with one of the generic verbs in a canonical complex verb construction. Simple and complex verbs both encode unitary events, defined as conceptual units that correspond to units of packaging in speech. In the case of complex verbs, the event may comprise subevents. However, the complex verb construction itself does not allow any conclusions about the semantic relationship between its constituents. In the remaining chapters, the meaning of verbs and coverbs, and their semantic relationship in canonical complex verbs, is investigated more closely. Chapter 4 is devoted to the argument structure of generic verbs, coverbs and complex verbs. According to the construction-based model that is employed, argument expressions (in Jaminjung both case-marked noun phrases and pronominal prefixes on the verb) are not directly predicted from the valency of the predicate, but are considered as constructions in their own right. In this approach, valency is defined in semantic terms as the number and type of semantic participants. Central participants can be operationally defined, for Jaminjung, as those participants which are encoded obligatorily and/or as core arguments (personal prefixes or absolutive noun phrases) in all uses of a given predicate. This criterion is particularly relevant in determining the valency of coverbs, since they usually do not enter argument structure constructions on their own, but only in combination with a verb. The semantic participants of predicates are linked to independently existing argument structure constructions on the basis of the compatibility of the participant’s role with the constructional meaning. This approach allows for a unified solution of a number of problems in the syntactic analysis of Jaminjung. These include the complex interaction of pronominal prefixes and case-marked noun phrases, the use of the same ‘ergative’ case to mark agents and instruments, and the representation of ‘agents’ not only by ergative-marked noun phrases, but also by noun phrases bearing a contrastive agent marker different from the ergative, and by unmarked (absolutive) noun phrases. In particular, the construction grammar approach allows for a simple and elegant representation of argument sharing in complex predicates: this can be described as the linking of one participant of a verb and a coverb, respectively, to a single morpho-syntactic argument expression. Thus, for example, the single participant of a monovalent coverb can receive very different expressions depending on whether it aligns with the single participant of a monovalent verb, the first participant of a bivalent verb, or the second participant of a bivalent or trivalent verb. In the account developed here, it is not necessary to postulate polysemy for the coverb in cases like these. Rather, the differences in argument structure result from the different contributions of the generic verb and the argument structure construction itself to the complex expression. In Chapter 5, the meaning and use of each of the generic verbs are examined in some detail in order to adduce evidence for the claim that the choice of a verb amounts to an act of categorisation, and that the verbs do not classify coverbs, but events. In other words, the categorising function of generic verbs is not restricted to their use in complex verbs. As simple verbs they describe an event of a specific general type, but may receive a more specific interpretation in context. As part of a complex verb they categorise the event encoded jointly by coverb and verb. Evidence for this view comes, first, from the fact that most coverbs combine with more than one verb, that is, the coverbs are not associated with a single ‘classifying’ verb (!). Second, even loans which are integrated into Jaminjung as coverbs are combined with verbs productively. Third, the generic verbs can generally be shown to have the same meaning both as simple verbs and as part of most complex verbs. This finding contrasts with the widely accepted view that in languages of this type, verbs, when part of complex verbs, are semantically empty and are mere carriers of verbal inflections, comparable to auxiliaries. This is not to say that verbs are never polysemous. Their secondary senses, if any, can however be related to their basic sense by regular semantic shifts due to processes such as metaphor, metonymy, and semantic bleaching (i.e. loss of semantic components). The main semantic components of the generic verbs – which are at the same time the features that are relevant for event categorisation in Jaminjung – include the number of salient participants in an event (reflected in the valency of the verb that is used), and the components of location, change of location and locomotion, contact and affectedness, transfer of an entity to, or away from, a third participant, and ‘internal cause’, and (to some extent) telicity. There are also verbs available to express ingestion, visual perception, and creation. For locomotion events, further distinctions pertain to the direction of motion and the involvement of a concomitant participant. For events of contact and affectedness, mere contact is distinguished from impact, and impact events are further subdivided according to the shape and trajectory of the agent/instrument making the impact (e.g. ‘with a pointed end’, ‘with an edge’, ‘with the foot’). Some types of affectedness by non-physical means as well as change of state, non-visual perception, and memory are covered by secondary senses of a number of verbs. The use of the verbs is, however, not motivated through their meaning alone, but also through the paradigmatic oppositions to other verbs that they enter into, and the application of general pragmatic principles. The resulting system of categorisation, while based on components which are often described as corresponding to ‘basic’ events, is quite language-specific. Chapter 6 deals with the meaning of the coverbs, and accounts for their use in complex verbs. Coverbs are divided into formal classes on the grounds of their possibility to combine with the same set of verbs. It is argued that coverbs belonging to the same formal class also form a semantically coherent set, and that these classes largely correspond to predicate classes found in other languages. In the case of the coverbs, too, the principle of monosemy is applied, such that the meaning of a coverb is taken to correspond to only those components that are present in all of its combinations with verbs. The unification of meaning components of coverbs and verbs in canonical complex verbs is then investigated for coverbs from each class. It turns out that the semantic relationships encountered in complex verbs can be of various types. The verb may be semantically included in the coverb, it may show partial semantic overlap with the coverb, or coverb and verb may not overlap at all semantically. In the latter case, the coverb is interpreted as expressing a phase of the event described by the verb, a manner in which it is performed, or a resulting event (where the causing event is expressed by the verb). With motion verbs, the reading can also simply be one of ‘associated motion’, i.e. motion simultaneous with, or followed by, another event. However, these differences should be seen as differences in semantic interpretation only; formally, all complex verbs instantiate a single type of complex verb construction. In Chapter 7, the findings of the previous chapters are summarised and placed in an areal and typological context. First, Jaminjung is compared to other Northern Australian languages with similar types of complex predicates. Comparative evidence is adduced for a diachronic scenario where larger verb systems get reduced to a small class of verbs which eventually gets extended again through complete lexicalisation of (previously) complex predicates. Second, the nature of coverbs, generic verbs, and complex verbs is examined from a cross-linguistic perspective. It is argued that coverb-verb combinations of the type found in Jaminjung and a number of other Northern Australian languages, although they bear many functional similarities with other types of complex predicates such as serial verbs, particle verbs or light verb constructions, should be considered a distinct type of complex predicate. Finally, it is pointed out that Jaminjung only ‘codifies’ in its lexicon and grammar a tendency that can be observed in many other languages, and perhaps even universally. In spoken, spontaneous discourse speakers tend to rely on a rather small class of semantically general high-frequency verbs, which may or may not be combined with other — often non-finite — elements. What is distinctive about Jaminjung and other Northern Australian languages is that the semantically most generic verbs with the highest frequency form a distinct part of speech and moreover participate in a system of overt event categorisation.