Thematicity in Lushootseed syntax

The surface form of sentences in Lushootseed is governed primarily by considerations of Communicative Structure, in particular Thematicity. This paper will argue that Thematicity, rather than part of speech, governs the selection of the syntactic predicate in Lushootseed, and will offer some preliminary formal treatments of the phenomenon in terms of ranked constraints governing the transition between SemR and DSyntR. It will also be shown that the typological distinction between Lushootseed and more familiar languages like English can be accounted for by variable rankings of a shared set of constraints on this transition. The surface form of sentences in the Salishan language Lushootseed, as in other languages in its family, is conditioned to a remarkable degree by Communicative Structure (Mel’!uk, 2001, a.k.a. Information Structure—Lambrecht, 1994; Vallduví, 1992). While previous attempts to come to terms with this aspect of Salishan syntax have made use of concepts such as Topic and Comment (Davis and Saunders, 1978) and Discourse Topic (Beck, 2000; Kinkade, 1990), this paper will attempt to account for a wider range of the effects of Communicative Structure on Lushootseed syntax by applying the model of Semantic Communicative Structure (Sem-CommS) outlined by Mel’!uk (2001). It will be argued that, contrary to traditional approaches to syntax, which give a priori primacy to lexical and syntactic categories in clause structure, Lushootseed requires that precedence be given to Communicative Structure in the organization of the clause. Some preliminary steps will also be taken towards formalizing a system of ranked constraints on the expression of particular elements of the Sem-CommS in Deep Syntactic Structure, and it will be shown that at least some of the typological differences separating Salishan from more familiar languages can be accounted for by the differential ordering of the same basic set of constraints. 1 Lushootseed clause structure Lushootseed, like other Salishan languages, uses a system of pronominal clitics, agreement-markers, and a fairly rigid VSO word-order to encode grammatical relations (Czaykowska-Higgins and Kinkade, 1998). The simple transitive clause is illustrated in (1): (1) a. !u–g"#$’–t $#d ti sq"#bay! PFV–seek–ICS 1SG.SUB DEF dog ‘I sought the dog’ 1 The abbreviations used in glosses are: 1,2,3 = first-, second-, third-person; ADD = additive; ATTN = attenuative; DEF = definite; DC = diminished control; DIST = distal; ICS = internal causative; INTJ = interjection; NDEM = non-demonstrative; PASS = passive; PFV = perfective; PL = plural; PR = preposition; PROX = proximal; REM = remote; REFL = reflexive; SG = singular; SUB = subject. b. !u–g"#$’–t $#% ti sq"#bay! PFV–seek–ICS 1PL.SUB DEF dog ‘we sought the dog’ c. !u–g"#$’–t $#x" ti sq"#bay! PFV–seek–ICS 2SG.SUB DEF dog ‘you sought the dog’ d. !u–g"#$’–t $#l#p ti sq"#bay! PFV–seek–ICS 2PL.SUB DEF dog ‘you guys sought the dog’ e. !u–g"#$’–t Ø ti sq"#bay! PFV–seek–ICS 3SUB DEF dog ‘he/she/it/they sought the dog’ (Hess, 1995: 10) Pronominal subjects are marked by one of a series of matrix subject clitics, the third person in this series being Ø and not making a distinction for number (Beck, 2000). A peculiarity of Lushootseed, shared to a certain extent by some other languages in the family (Gerdts, 1988; Kinkade, 1990), is that transitive sentences with both an NP subject and an NP object are disallowed (Hess, 1973). Sentences with thirdperson subject and object undergo obligatory pronominalization of the subject, surfacing as sentences like (1e). An interpretation of (1e) where the NP following the verb is the subject/AGENT (i.e., ‘the dog sought him/her/it/them’) is disallowed by what Gerdts (1988) refers to as the One-Nominal Interpretation Law. Interpretation of transitive clauses in discourse is facilitated by a reference-tracking system built around a strong constraint that subjects be topical (Beck, 2000; Kinkade, 1990). In contexts where the identity of the subject of a transitive clause is not recoverable from discourse, or where both event-participants must be specified for communicative reasons, the passive voice is used: (2) !u–g"#$’–t–b !# ti $’a$’as ti sq"#bay! PFV–seek–ICS–PASS PR DEF child DEF dog ‘the dog was sought by the boy’ (Hess, 1995: 23, ex. 6a) Like the English passive, the passive in Lushootseed promotes the direct object to subject (Sub) and demotes the active voice subject to an oblique agentive complement (AgCo) phrase, introduced by a preposition (!"). The order of arguments can be either Sub >> AgCo or, as shown here, AgCo >> Sub, the order in (2) being more prevalent. One of the more remarkable characteristics of Lushootseed syntax is the flexibility it displays with respect to which parts of speech are eligible syntactic predicates (Beck, 2002). As in most Salishan languages (Kinkade, 1983), words corresponding to English verbs, nouns, adjectives, adverbs, and demonstrative pronouns are all potential predicates, as in the nominally-predicated expressions in (3): (3) a. !aci%talbix" $#d Indian 1SG.SUB ‘I am an Indian’ (Hess & Hilbert, 1976: vol. I, 36) b. s!uladx" ti!i% salmon DIST ‘that is a salmon’ (Hess & Hilbert, 1976: vol. I, 7) Non-verbal predicates like !aci"talbix# ‘Indian’ in (3a) take the same subject inflections as do the verbal predicates in (1) above; copular constructions with NP or demonstrative subjects like that in (3b) simply juxtapose subject and predicate, the latter appearing in sentence-initial position. Unlike nouns and other 2 The translation “boy” is from $’a$’as ‘child’ and the non-feminine determiner, ti; ‘girl’ would be tsi $’a$’as. nominal elements used as arguments, nominal predicates do not take determiners unless the sentence identifies the subject with a specific individual. Non-verbal predicates are not confined to simple expressions of identity like those in (3); constructions like that in (4), with a nominal predicate and a complex nominal acting as subject, are quite routine: (4) wiw’su ti!#! !u–$ala–d ti!#! sq"#bay! children PROX PFV–chase–ICS PROX dog ‘those who chased the dog are the children’ (Hess, 1995: 99) The syntactic predicate in (4) is the bare noun wiw’su ‘children’, while the subject is a relative clause headed by the determiner ti!%! (Beck, 2002). It is the predicative use of nouns in constructions like those in (3) and (4) which have led some researchers (e.g., Jelinek & Demers, 1994; Kinkade, 1983; Kuipers, 1968) to argue that there is no distinction between nouns and verbs in Salishan languages. While this is probably an over-reaction, it is the case that part of speech plays less of a role in the selection of the syntactic predicate of a sentence than it does in most other languages, where there is a strong tendency for sentences to have verbal predicates. Instead, predicate selection in Lushootseed depends crucially on Communicative Structure. 2 Effects of Thematicity on syntactic structure The component of Sem-CommS that has the greatest effect on Lushootseed syntax is Thematicity, which is the driving force behind the selection of the syntactic predicate. A similar effect is discussed for another Salishan language, Nuxalk (Bella Coola), in Davis & Saunders (1978). In Davis & Saunders’ terms, the structure of a Nuxalk clause involves a bi-partition between the part of the sentence that is the “Comment” and that which is “Topic.” In our terms, this translates into the Nuxalk sentence being organized so that the Sem Rheme is expressed as the syntactic predicate and the Sem Theme as its subject. Beck (1997) argues that Lushootseed has the same pattern, as shown by question-and-answer pairs such as that in (5): (5) a. !u–!#x &id k"i ki–k#wi$ PFV–what.happen REM ATTN–hunchback ‘what happened to Little Hunchback?’ — !u '’al’ b#=!u–sax"#b–dx"–but ti!i% ki–k#wi$ INTJ also ADD=PFV–run–DC–REFL DIST ATTN–hunchback ‘oh, Little Hunchback also managed to escape’ [DM Basket Ogress, lines 79–80] The question in (5a) is a narratively-focused question asking about an event in which a particular Thematic event-participant is involved, and elicits a narratively-focused response with a Rhematic verbal predicate. The question in (5), however, asks for the identity of an unknown participant in a Thematic event, this event being expressed as a headless relative clause in subject position of a sentence whose predicate is the interrogative word stab ‘what?’. The response mirrors this structure exactly, substituting the requested information for the interrogative, giving us a sentence with a Rhematic nominal predicate. Subsequent work has found that this pattern also occurs more generally in Lushootseed narrative and other discourse contexts where the event is Thematic (and, generally, Given) and the Rheme is an eventparticipant or some other non-verbal element of the sentence (Beck, 2000; Beck, 2002). This is an interesting situation from a theoretical point of view in that traditional approaches to syntax generally approach clause structure as being built around phrasal projections of lexical elements whose part of speech (or the projections of the functional/inflection categories associated therewith) determine whether particular elements are realized in predicate or argument positions. Thematicity in such approaches tends to be 3 The position that Salishan languages do not distinguish nouns and verbs is not the current consensus position held by most specialists in these languages—cf. van Eijk & Hess (1986), Kroeber (1999), and Beck (2002). 4 Materials not cited as being from published sources are drawn from unpublished texts kindly provided by Thom Hess; these are referred to by speaker’s initial followed by title of the text and line number(s). treated as a secondary phenomenon with less import for clause structure; however, data like that in (5) shows that Thematicity plays a much more fundamental role in Salishan languages, and suggests that it might be worthwhile to re-examine some of these traditional assumptions. In the following section, I will illustrate how the effects of Thematicity on syntactic structure can be modeled from the perspective of text generation, using some of the formalisms from Meaning-Text Theory, beginning with the selection of the syntactic predicate (Section 2.1), and then moving on to constraints on grammatical voice (2.2). 2.1 Constraints governing the selection of syntactic predicates From the point of view of text synthesis, the initial step in mapping between SemR and DSyntR is the selection of what is called the entry node, that semanteme (or configuration of semantemes) in the SemR that will be lexicalized as the top node of the DSyntS (the matrix predicate). In the SemR in Figure 1 above, for example, the entry node is ‘seek’, which is lexicalized as the verb g#%$%d, the matrix predicate in the DSyntR. One of the reasons that ‘seek’ is chosen is that it is the Communicatively Dominant Node in its sub-network of the SemR. In this context, Communicatively Dominant means that the sub-network of the SemS ‘past’ ! ‘seek’ ! ‘dog’ can be reduced to ‘seek’ without changing its referent—that is, its referent is a seeking event rather than, say, a dog (if ‘dog’ were the Comm-dominant node in this configuration, the corresponding DSynt tree would be a relative clause, ‘the dog that was sought’). The second consideration that determines the selection of ‘seek’ as the entry node is the part of speech with which its meaning is lexicalized. In English, ‘seek’ corresponds to the meaning of a verb, LOOK(FOR), which is selected as the top node of the DSyntS of the boy sought the dog. In most languages, part of speech is the primary factor governing the selection of entry nodes: by far the majority of the world’s languages select as entry node semantemes that are semantic predicates and whose most natural expression is a verb. However, Lushootseed departs from this pattern in preferring elements that are Rhematic, whatever their lexical class, allowing for a wide range of non-verbal predicates. Procedural rules for the selection of entry nodes in languages like English, Russian, and French are set out in work by Iordanskaja and Polguère (Iordanskaja, 1990; Iordanskaja & Polguère, 1988) and discussed in the context of Sem-CommS by Mel’!uk (2001). Without getting bogged down in technical details, the relevant rules proposed by these authors can be restated as the ranked constraints in (6): (6) Semantically-Predicative Entry Node (SPEN) The entry node is a semantic predicate. Verbal Entry Node (VEN) The entry node is most naturally lexicalized as a verb. 5 A more detailed discussion of Comm-Dominance and methods for determining the Comm-Dominant node can be found in Iordanskaja and Polguère (1988) and Mel’!uk (2001: 29ff.). 6 This, of course, may depend on language-specific characteristics of the lexicon such as the (non-)existence of verbal expressions for certain semantemes (the case in point being the absence of a verbal expression of ‘be’ in Lushootseed—see below). The term “most naturally” is deliberately vague, allowing for lexical and other types of idiosyncrasies to override more rigid considerations such as the existence in the lexicon of a direct expression of a particular semantic predicate (viz. the case of the English expression BE HUNGRY which is the most common expression of the stative predicate ‘hunger’, in spite of the existence of the verb HUNGER). ‘dog’ ‘boy’ ‘seek’