Serial Verb Constructions in Typological Perspective

Serial verb constructions are a grammatical technique whereby two or more verbs form one predicate. A sequence of verbs qualifies as an SVC if there is no marker of syntactic dependency between the components (and, in addition, for languages which distinguish between finite and non finite verbs, neither component can take a separate nonfinite marking; the whole construction has to be nonfinite, as in example (9), from Lango). SVCs are distinct from idiomatic double verb sequences which have restrictions on their mood, tense and aspect choices (as in European languages). SVCs form one prosodic unit (see §2.3). An SVC describes what is conceptualized as one integrated situation, or one event. Semantically, such an event may be composed of a series of sub-events. 'Single-scene' SVCs correlate with cohesive, tightly-knit structures with shared participants; they tend to be more fused in their surface realization than 'multiscene' SVCs. These correlate with less cohesive, less tightly bound constructions, and may even be reminiscent of clause sequences. The differences can be accounted for by the principle of iconicity in grammar. All serializing languages have same-subject SVCs. Prototypical SVCs share all arguments. Lack of argument sharing is associated with less cohesive and less tightly-knit structures. Event-argument SVCs are a type of SVC with no shared arguments. The event or state denoted by one component is predicated on the entire situation referred to by an SVC. By their composition, SVCs fall into two broad groups. Asymmetrical verbs consist of a 'minor' verb from a closed class, and a 'major' verb (the head of an SVC) from an open class which determines the transitivity of the whole construction. The minor verbs tend to grammaticalize into markers of direction, aspect, and valency changing (see §3.4). Symmetrical SVCs consist of components chosen from major lexical classes. They do not have a head, and tend to give rise to lexical idioms. Languages with a grammaticalizing tendency may, synchronically, have no asymmetrical SVCs (as is the case in Ewe) . Languages with a lexicalizing tendency may have no symmetrical SVCs (as is the case in Tetun Dili). Productively serializing languages tend to have SVCs of both kinds, while languages with limited serialization have just asymmetrical SVCs. The distinction between asymmetrical and symmetrical SVCs may be viewed as a continuum, depending on the semantic and functional overlap between subtypes of both, and on the corn position of closed and open classes of verbs. Serial verb constructions can be contiguous or non-contiguous. They may form one grammatical and/or phonological word, or be multi-word. In multiword SVCs, various grammatical categories can either receive concordant marking (on every component) or be marked just once. The person of the subject is more likely to receive concordant marking than any other category. SVCs of all types and structures show the same functional and semantic properties and tendencies. The present framework- inclusive in character- allows us to apply the proposed parameters to SVCs in a wide variety of languages (overcoming some terminological traditions, such as an Africanist tendency to consider only multi-verb SVCs as SVCs, and discarding one-word constructions, as found in Igbo, and also Olutec, Tariana, Lakota, Yimas, and others ). Coexisting types of SVCs in a single language differ as to whether they have concordant or single marking for the various categories discussed here. Synchronically, if there are several types of SVCs in one language, they are likely to be independent grammatical processes, each with a grammaticalization path of its own, and each used to convey a different type of grammatical meaning. SVCs could be conceived of in terms of a multidimensional continuum, covering such parameters as the possibility of pause marking (see §2.3 ) , of semantic cohesion and eventhood (see §2.5 ) and historical development (or grammaticalization ). Verb serialization is a syntactic resource which allows the speaker to express various aspects of a situation as a single cognitive package within one clause and with one predicate. Such a cognitive packaging strategy is highly diffusable- and thus verb serialization is typically a property of a linguistic area. If a language has no or little bound morphology, it is particularly likely to develop multi-word verb serialization, although synthetic languages are not immune to similar developments. SVCs show semantic and functional (rather than formal) similarities with other multiverb constructions, both monoclausal- such as converb constructions and clause-chaining (see Chapter 15, this volume)- and biclausal- such as coordinate and overlapping clauses in Ewe (Chapter 5). These similarities justify considering SVCs as part of a multidimensional continuum of multiverb structures. Diachronically speaking, links can be established connecting focal points on this continuum (so, for instance, a special marker of SVCs, as in Khwe and Yimas, indicates that these constructions come from multiverb structures of a different, non-serial, kind). Despite the considerable literature on verb serialization, much remains to be investigated in order to obtain a further cross- linguistic perspective on its varied facets. Some of such issues- which should be analysed from both a structural and a semantic viewpoint- include: -the semantic and pragmatic functional motivation for optional verb serialization; -the semantics and pragmatic functions of optional concordant marking of grammatical categories; -further analysis of various origins and grammaticalization paths for different kinds of SVCs; -further analysis of several coexisting SVCs where they occur in a single language; -the cognitive and conceptual correlates of verb serialization, as a focal point within a continuum of multiverb constructions.