The paper proposes rules for linear placement of Serbian second-position clitics under text synthesis from the viewpoint of the Meaning-Text theory. Two specific questions are addressed: syntactic vs. prosodic factors in clitic placement and the notion of constituent in the Meaning-Text framework. The proposed description is compared with those within the Chomskian paradigm.
A Meaning-Text Model of language L is a stratificational dependency-based model. Clitic placement rules are part of its Surface-Syntactic module, which maps a Surface-Syntactic Representation [= SSyntR] of a clause to its Deep-Morphological Representation [= DMorphR], i.e., performs the linearization, prosodization and morphologization of the SSyntR. The basic structure of the SSyntR is an unordered dependency tree; that of the DMorphR is a fully ordered string of wordforms, including the clitics. There are no constituents in the SSynt-tree: constituents are obtained as a result of the linearization/prosodization of SSynt-subtrees and appear in the morphological string. Clitic linearization rules build the clitic cluster and position it after an appropriate constituent [= host] in the morphological string, the latter operation involving the identification of potential hosts and the choice of the actual host.
Unlike full-fledged words, clitics are not linearly positioned with respect to their syntactic governors: in the SSynt-tree, different clitics have different governors (pronominal clitics depend on the lexical verb, the emphatic particle LI depends on the lexeme it ‘emphasizes’, etc.) or are governors themselves (an auxiliary clitic is the top node of its clause), but in the morphological string they are positioned together, as a cluster, after a constituent which need not (and most often does not) correspond to the SSynt-governor of any of the clitics.
Properties of a constituent relevant for determining its hosting capability are prosodic (heaviness, i.e., the number of stressed wordforms, the necessity of a pause after it), communicative (e.g., contrastiveness) and syntactic (syntactic role and/or composition); they are (partially) generalized as features [±heavy], [±detached], [±autonomous] and [±contrastive]. According to these features, constituents are identified as non-hosts or potential hosts, the latter being subdivided into skippable vs. non-skippable hosts and insertable vs. non-insertable hosts. A constituent followed by an obligatory pause is marked as [+detached], which makes it into a non-host, a constituent featuring at least two stressed wordforms is [+heavy] and a skippable host, a constituent with the feature [+heavy] and the appropriate syntactic composition (e.g., having the form ‘Premodifier+X’) is an insertable host, an so on.
Two basic clitic placement rules specify possible landing sites for the clitic cluster.
Rule-1: Clitic cluster is placed after the first non-skippable host of its clause or after any of the skippable hosts preceding the non-skippable host.
Rule-2: Clitic cluster placed after an insertable host may be inserted into this host, after its first stressed wordform
In the sentence below, featuring three skippable hosts, two of which are insertable, and two non-skippable hosts, there are six possible landing sites (symbol ¯) for the clitic auxiliary su:
[U tim ¯ okolnostima]skip, insert ¯ [ideje Čomskog]skip ¯ [dalju ¯ razradu]skip, insert ¯ [doživele]non-skip¯ [u delu …]
The actual landing site of the cluster is determined by preference rules for skipping and insertion.