In this talk I examine non-verbal predication in Karuk, an endangered indigenous language of northwest California. Karuk is thought to be part of the Hokan stock, and share the polysynthetic character of the neighboring northern Hokan languages. The example in (1) shows a typical Karuk sentence, in which the verb contains several derivational affixes, expressing event iteration (ip-), purpose (-ar), and direction (-uk; -vu), and several inflectional affixes, expressing participant number and person (na-), aspect (durative -tih) and tense (future -heesh):

(1) hâari vúra xasík na-pi-mus-ar-ûuk-vu-ti-heesh
sometime INTENSIVE then.FUT 2SG:1SG-ITER-see-PURP-hither-to-DUR-FUT
`You can come back to see me sometime.’
Julia Starritt, “Coyote Marries His Own Daughter”

While verbal predication is the norm, Karuk also exhibits nonverbal predication, where a noun, adjective, adverb or quantifier functions as the predicate of the clause. As shown in (2), non-verbal predication exhibits tense marking, but not agreement:

(2) Náa vúra yâamach-heech
1SG INTENSIVE pretty-FUT
`I’m going to be pretty’
Imkyanváan, “Coyote Doctors a Girl”

The suffix -hi derives verbs from adjectives and nouns and the resulting forms do agree, as the minimal pair in (3) shows. (The sentences in (3) occur in close succession in Emily Donohue’s telling of “The Pikiawish at Katamin”.)

(3)
a. xás tá-kóo pa-’ir
then PERF-all the-world.renewal.ceremony
`Then the world renewal was over.’

b. yáas u-kôo-hi-ti pa-’ir
then 3SG-all-VBLZ-DUR the-world.renewal.ceremony
`Then the world renewal ended.’

In (3a), the perfect marker tá attaches to the quantificational root koo `all’. There is no agreement on the resulting form. In (3b), the verbalizing suffix -hi attaches to the same root, koo, followed by the durative suffix -ti, and the resulting form bears 3SG subject agreement (u-).

Bright’s (1957) interpretation of this pattern is that (2) and (3a) are instances of non-verbal predication. The tense and aspect markers that cooccur with non-verbal predicates are clitics, and hence not limited to verbal hosts, whereas the agreement markers are affixes that can only attach to verbs. Macaulay (1989) offers a reanalysis wherein tense and aspect marking of seemingly non-verbal predicates always involves verbalizing by -hi, though regular morphophonemic processes conspire to obscure the presence of this morpheme on the surface. Macaulay’s analysis explains why all non-verbal predicates take h-initial allomorphs of the suffixal/enclitic tense and aspect markers, something that Bright must stipulate. To account for the lack of agreement on verbs derived in this manner, she appeals to semantics (p. 176-8), specifically that agreement is absent due to such clauses having an equative meaning.

I evaluate Macaulay’s claim against a body of examples drawn from published texts and from original field work and conclude that it is not supported. I tentatively suggest an alternative analysis, in which -h is a stem extender that readies a non-verbal element for tense and aspect suffixes, but does not change its part of speech. That is, in effect, a return to Bright’s analysis, though with a possibility of incorporating Macaulay’s verbalizing analysis as a diacronic source of the stem extending -h.

In the latter part of the talk, I place the observed morphosyntactic patterns of word class, tense-aspect marking and agreement in a typological (Stassen 1997) and a theoretical (Baker 2003, 2008) perspective, and discuss possible syntactic differences, in Karuk, between clauses involving non-verbal predication and clauses involving verbal predication.

Given the recent explosion of work on agreement within the frameworks of Minimalism and Distributed
Morphology (Béjar 2003, Baker 2008, Bobaljik 2008, Béjar & Rezac 2009, Preminger 2011),
the typology of agreement and certain noncanonical agreement patterns have come into theoretical
focus. This talk is concerned with one such noncanonical pattern, discontinuous agreement (DA),
which involves a deviation from the expected one-to-one relation between (sets of) features and
their morphological expressions. For instance, (1) is an example of coreferential DA in which multiple,
coreferring agreement features are encoded by distinct morphs. Combinatorial DA, shown
in (2), involves a single feature category (here, number) splitting into more than one component
value (nonsingular, which is used in both dual and plural forms in this language, and plural in the
strict sense), each of which is realized by a separate morph but all of which are required to yield
the intended meaning.

(1)
zuek z-atoz-te
2PL 2-come-PL
Basque: ‘You (≥2) come.’

(2)
do:-ya:-di-l-yo’
NEG-PL.STRICT-1NSG-CLS-love
Hupa: ‘We (≥3) do not care for it.’

The present study makes three contributions to the understanding of agreement. First, it establishes
DA as a robust phenomenon through a systematic study of DA in 40 genetically and
geographically diverse languages; second, it offers new generalizations about the behavior of agreement
features (person, number, and gender) in discontinuous agreement; third, it provides an
analysis of agreement in terms of two-dimensional phi-sets and cyclic insertion that captures the
noncanonicality and observed patterns of DA.

Generative analyses within the Minimalist framework vary widely in how they account for case assignment in so-called ergative languages. Legate (2008) proposes that for a large class of ergative languages, what is taken to be Absolutive case is actually two cases with the same morphological form – Nominative case on sole arguments of intransitives and Accusative case on patients of transitives. Ergative case is inherent, assigned by a functional head (little v) in transitive clauses. I extend this analysis to Samoan and show that it correctly predicts a split in the distribution of the morphologically null case (traditionally termed Absolutive) – it is available as Accusative case for transitive patients in nominalised clauses (1), but not for the sole argument of intransitive nominalised verbs (2).

(1) e leaga lo’u iloa [mea lena]ACC
PRES bad my know thing that
“I don’t want to know about that” (lit. “my knowing that is bad”)

(2) e `ese lava le agi *(a) le matagi
PRES extraordinary the blow GEN the wind
“The wind is blowing hard.” (lit. “the blowing of the wind is extraordinary”)

Furthermore, as the null case is assigned by two distinct sources under Legate’s system (by T as Nominative and by transitive v as Accusative), the possibility of transitive clauses in which the null case is assigned to both the agent and patient is correctly predicted.

(3) sā fa’amoe [le tinā]NOM [lana pepe]ACC
PAST CAUSE.sleep the mother her baby
“The mother put her baby to sleep.”

This model of case assignment has implications for a theory of subjecthood in Samoan. The argument which may receive Nominative case (that is, transitive agents and sole arguments of intransitives) also passes some syntactic tests for subjecthood, for example, it is the argument dropped in control and raising constructions. This research is a preliminary step in a discussion of how case assignment and argument realisation properties converge in ergative languages.

References
Legate, Julie. 2008. Morphological and Abstract Case. Linguistic Inquiry 39(1): 55-101
Mosel, Ulrike & Evan Hovdhaugen. 1992. Samoan Reference Grammar. Oslo: Scandinavian University Press.

Abstract: In this talk I will argue that, in typical nominative-accusative languages, nominative case is not assigned by finite T or by agreement with any functional head. Rather, even in prototypical subject position it is the default case, showing up when the conditions for the assignment of all other cases are not met. The appearance of a dependency is due to a confusion of conditions on nominative case with conditions on the overtness of subjects. Conditions on overtness and case-marking tend to coincide, but a careful look at the data shows that they are orthogonal to each other, and that the latter has more to do with conditions on subject (in)dependence.(based on joint work with Sandhya Sundaresan)

Several properties of noun phrases in Moro (Kordofanian; Sudan) follow from an analysis involving movement of a nominal head to a position in [Spec, D], along the lines proposed by Matushansky (2006, LI 37.1). These include a) N-Dem-Num-Adj word order b) the syntax of adpositions, which intervene between nouns and demonstratives, and c) a definite clitic which only occurs overtly in the presence of demonstratives, possessives, and relative clauses. Evidence against the standard analysis involving head-adjuction comes from the ability of Moro nouns to strand their modifiers in the passive. I also adopt Matushansky’s proposal that the phonological effects of head-movement arise due to a separate post-syntactic operation, m-merger, which involves the fusion of two adjacent heads into a single phonological word. M-merger can account for the complementarity between the presence of the definite clitic with modifiers and definite interpretations of bare nouns, reducible to a case of allomorphy.

Adger and Ramchand (2005) claim that Scottish Gaelic relativization does not involve movement but only a static Ā-dependency. Their discussion leaves out an important relativization strategy which targets prepositional phrases, and which provides morphosyntactic evidence for movement, as it involves fronting of the preposition to precede the relative complementizer, as in (1).(1)
gloinne anns an do dhoirt thu am bainne ____
glass in.DEF AN pour.PAST.DEP you the milk
‘a glass into which you poured the milk’
(Andrew Dunn, p.c.)

(2)
am bòrd a bha an leabhar fodha ____
the table A be.PAST.IND the book under.3MSG
‘the table the book is under’
(Adger & Ramchand 2006: 10)

There are several striking characteristics of this relativization strategy. First, the preposition appears at the front of the relative clause, rather than being left in situ. Second, the fronted preposition obligatorily takes ‘definite’ inflection, regardless of the definiteness of the pivot. Third, the form of the complementizer an is unique to this preposition-fronting context.

I analyze the pattern in (1) as involving movement, and argue that the form of the preposition provides us with morphosyntactic evidence that there is a D° head associated with the relative clause (cf. Bianchi (1999), Borsley (1994)). In this talk I focus on the morphological derivation of the ‘definite’ inflection on the preposition, as triggered by the relative D° head, and thus attempt a unified account of ‘definite’ inflection in Scottish Gaelic.

References
Adger, David and Gillian Ramchand. 2005. ‘Merge and Move: Wh-Dependencies Revisited’ in Linguistic Inquiry 36: 161-193.
Adger, David and Gillian Ramchand. 2006. ‘Dialect Variation in Scottish Gaelic Relative Clauses’ in Rannsachadh na Gaidhlig, ed. W. McLeod: 1-15.
Bianchi, Valentina. 1999. Consequences of Antisymmetry: Headed Relative Clauses, Studies in Generative Grammar 46. New York: Mouton de Gruyter.
Borsley, Robert. 1994. ‘Relative Clauses and the Theory of Phrase Structure’ in Linguistic Inquiry 28: 629-647.

True clitic doubling involves multiple expressions of a single argument in different structural positions. It involves configurations in which a phonologically bound morpheme (the clitic) expresses features of a full noun phrase in an argument position (the associate). True clitic doubling has traditionally been argued to arise via agreement, so that the clitic is a manifestation of an agreement relation between a verb and the associate (see Borer 1984, Jaeggli 1986, Suñer 1988). An alternative possibility is that the clitic is a nominal (e.g. a pronoun of category D) related to the associate via movement (see Sportiche 1996 and Anagnostopoulou 2003); clitic doubling, then, involves the simultaneous realization of both the head and the foot of a movement chain.

Here I argue, on the basis of evidence from Bulgarian, for the latter analysis, showing that Bulgarian clitic doubling has the properties of movement–i.e. it does not involve agreement, as is standardly assumed for this language (Rudin 1997, Franks & King 2000, Pancheva 2005). I provide support for this claim by considering a number of diagnostics which distinguish between clitics that reflect agreement processes and clitics that do not, adding to a growing body of recent literature on the issue (e.g. Kramer 2011, Nevins 2010, Preminger 2009). Specifically I argue that the clitic is a reduced articulation of the higher copy of a raised complement.

I investigate data discussed by Pullum (1987) and Pesetsky (1991), in which the pronominal argument of a verb appears to take as its referent the content of a conditional clause:

1. I would like it if the tacos were tasty.

Two readings are possible for the pronoun in (1). Under the “normal” reading, it may indicate any contextually salient referent—say, the taqueria the speaker is in. Under what I term the “adverbial-referring” reading, the pronoun appears to take the content of the adverbial clause as its referent: If the tacos were tasty, the speaker would like that that tacos were tasty. The adverbial-referring reading is associated with a number of anomalies not associated with the normal reading:

2. Lexical restrictions on the main-clause predicate
*Gretchen would demand it if the fog lifted.

3. Main-clause predicate effects on NPI-licensing
*Max would love it if anyone were to touch the painting.

4. Choice of anaphor
*This bothers me if he yells.

Previous analyses account for these facts via a movement relation between the adverbial clause and the argument position, and require some sort of novel morphosyntactic or semantic operation. I show that these analyses cannot capture the entire range of data. I propose a relatively simple syntactic analysis, in which the adverbial is a typical adjunct and is mapped into the restrictor of a tripartite quantificational structure; the pronoun is a typical pronoun, which happens to refer to the content of the adverbial clause in the adverbial-referring reading. The effects seen in (2)-(4) are shown to fall out from the interaction of quantification, reference, and the lexical properties of predicates—and do not require novel operations in the grammar.

Infinitives in both European and Brazilian Portuguese (EP and BP, respectively) can be marked for agreement, a verb form known as the
inflected infinitive:(1) Eu persuadi ele a esperar.
I persuaded.1sg he to wait.INF.3sg

(2) * Eu persuadi ele a esperarem.
I persuade.1sg he to wait.INF.3pl

(3) Eu persuadi eles a esperarem.
I persuade.1sg they to wait.INF.3pl

This form coexists in the language with a noninflected infinitive; importantly, overt subjects in infinitival clauses always trigger agreement, and can be assigned nominative Case by the inflected infinitive.

Although a large class of verbs can take both forms in their complements, there seem to be both semantic and syntactic differences between the alternatives. Focusing on two verbs, ver (see) and mandar (order), I make three major claims about these complements. First, (1) I demonstrate that there is a clear surface contrast between the inflected and noninflected infinitives. Their external arguments have different case marking — with the former we see only the nominative, with the latter, only the accusative. Furthermore, a number of syntactic tests show that (2) the accusative-marked arguments are part of the matrix clause, whereas the nominative-marked are part of the embedded clause. Additional tests relating to the availability of certain items inside the complements indicate that (3) the constituent containing the inflected infinitive is fully clausal, whereas that containing the noninflected infinitive seems to be smaller.

I will then show a set of facts that are currently unaccounted for and seem inconsistent with this third claim; namely, that the subjects
(but not the objects) of inflected infinitive complements seem to be in the same binding domain as the subjects of the matrix verbs that
select them, challenging the idea that there is a clause boundary between them.

Verb serialization and nominalization are two prominent phenomena in descriptive and theoretical syntax. In this talk I raise a number of issues that result from the interaction between these two widely attested phenomena in the literature: nominalization (e.g. Chomsky 1970) and verb serialization (e.g. Baker 1989, Bodomo 1993, Collins 1997). Based on data from Dagaare, a Gur language of West Africa, this paper analyses a serial verb construction which is a type of complex predicate construction in which all the verbs in a series are nominalized, with only one of the verbs carrying the nominalization affix (Bodomo and Oostendorp 1993, Bodomo 2004, Hiraiwa and Bodomo 2008). Such a rare complex predicate construction is then the basis for renewed questions about the nature of complex predicatehood, diathetic syntactic alternations, and lexical categorial differences involving nouns and verbs across languages. In the talk, I propose a syntactic representation of these nominalized serial verbal predicates in which the verbal predicates are basically interpreted as VPs headed by a nomP functional projection. Semantically, we propose that nominalized serial verbs, like their purely verbal counterparts, express a complex event. It is thus concluded that while verbal and nominal predicates obtain from the same minimal constructs, the difference between pure serial verbs and nominalized serial verbs is due to the fact that a semantic feature, [+nom], parallel to the syntactic functional projection, nomP, imposes nominal features on the whole complex. This analysis is extended to complex verbal constructions in English.

Acehnese (Malayo-Polynesian; Sumatra, Indonesia) is a language which instantiates an Indonesian-style voice-marking system wherein a particular pivot nominal is responsible for bearing many of the canonical properties of “subjecthood” (Chung 1976). In this talk, I show that this system is Indonesian-like (as opposed to the Tagalog or Philippine-type) insofar as it has three thematic voices: (i) an active/agent voice, (ii) an object voice, and (iii) a more canonical English-style passive voice (Legate 2008, to appear). I then turn to contrasting three main theories of Austronesian voice via the lens provided by Acehnese: the classic two-subjects analysis of Guifoyle, Hung, and Travis (1992); the phase-theoretic approach to Tagalog of Rackowski and Richards (2005); and the ergative-analysis of Aldridge (2005, 2006, to appear). I show that the data from agreement, word order, and variable binding are consistent with all the approaches, but that data from wh-extraction is suggestive of a split-ergative approach to Acehnese.