ComSyn Lectures

Speaker: Andrés Saab (Universidad de Buenos Aires / Consejo Nacional de Investigaciones Científicas y Técnicas)

Title: Ellipsis as silent doubling

Date: Thursday 6 April

Venue: Lipsius 30

Time: 15.15-16.30 hrs

Abstract

I address the well-known problem of the identity condition on ellipsis (semantic or syntactic?) from a novel (or maybe better, broader) perspective, namely, I contend that the identity problem forms a natural class with the also well-known doubling problem (Kayne/Jaeggli’s observation).

(1)        Juan     la         vio       a          María.

J            CL       saw     DOM   M.

Given a doubling configuration like this, there are questions at the basic level both from a syntactic (Case) and a semantico-pragmatic perspective (binding and so on). My point is that ellipsis consists of a silent doubling structure where a sometimes silent pronominal element (R_E in 2) and a full-fledged syntactic structure (E-site in 2) form a doubling configuration:

(2)        John saw someone but I don’t who [_RP R_E [E-site]]   Surface anaphora

 In contradistinction with an overt doubling like (1), in (2) R_E is in charge of the semantic resolution of ellipsis (as in many LF-copy approaches, although we will see essential differences) and the E-site is entirely resolved in the syntax via lexical-syntactic identity. Lexical-syntactic identity is simply an instruction for creating syntactic fossils. In other words, ellipsis applies in the syntax and instructs both interfaces for not realizing the E-site. A syntactic fossil, then, is just dead syntax. Put differently, ellipsis is conceived here as both PF and LF “deletion” (deletion = absence of interface legibility). “Ellipsis as doubling” has many empirical consequences. It provides a plausible solution to at least the following ellipsis puzzles:

(3)        a. Vehicle Change effects (without appealing to VC),

1. Indexical switches (without accommodation),
2. Expressive mismatches (under perfect syntactic identity),
3. Bias Vehicle Change also under perfect syntactic identity (i.e., legitimate ellipses where there is a mismatch in the bias of some lexical or functional item included in the E-site)

For each of these cases, I propose a derivation where antecedents and E-sites are syntactically identical (in a sense to be defined during the talk) without obtaining any semantic mismatch at LF. This is so because of two basic reasons: (a) the E-site is an LF fossil as already mentioned, and (b) R_E semantically resolves the ellipses by only making reference to the at-issue content of the antecedent. The two essential assumptions that are the core of this proposal receive independent corroboration from different empirical domains. As for R_E, Elbourne (2008), Bentzen et al (2013) and Messick et al (2016) have provided robust evidence for the presence of a pronominal (deep) element in surface anaphora. In this talk I will discuss two pieces of evidence in favor of the syntactic and semantic activity of R, namely, Inheritance of Content (Chung et al 1995 and Romero 2008, among others) and Barros’ effect (Barros 2012). As for LF “deletion”, our second core assumption, I will discuss evidence coming from the so-called allosemy view (Marantz 2013 and Harley 2014, among others), according to which lexical and functional items also require late insertion at the LF interface under particular locality conditions.

References

Bentzen, Kristine, Jason Merchant, and Peter Svenonius. 2013. Deep properties of surface pronouns: pronominal predicate anaphors in Norwegian and German. Journal of Comparative Germanic Linguistics 16:97–125.

Barros, Matthew. 2012. Else-modification as a diagnostic for pseudosluicing. URL

http://ling.auf.net/lingbuzz/001761, ms.,RutgersUniversity.

Chung, Sandra, William Ladusaw, and James McCloskey. 1995. Sluicing and Logical

Form. Natural Language Semantics 3:239–282.

Elbourne, Paul. 2008. Ellipsis sites as definite descriptions. Linguistic Inquiry 39:191–220.

Harley, Harley. 2014. On the identity of roots. Theoretical Linguistics 40: 225–276.

Marantz, Alec. 2013. Locality domains for contextual allomorphy across the interfaces. In Matushansky, O. and A. Marantz (eds.) Distributed Morphology today. Morphemes for Morris Halle. Cambridge, Mass.: MIT, press, 95-116.

Messick, Troy, Andrés Saab & Luis Vicente. 2016. Deep properties of a surface anaphora. Ms., UConn, CONICET-UBA, University of Potsdam.

Romero, Maribel. 1998. Focus and reconstruction effects in wh- phrases. Doctoral Dissertation, University of Massachusetts, Amherst.

Speaker: Beste Kamali, Universität Bielefeld

Title: Negative concord in Turkish polar questions

Date: Thursday March 23

Venue: Lipsius 30

Time: 15:15 – 16:30 hrs

Abstract 

Turkish is a strict negative concord (NC) language. That is, it has n-words rather than existential polarity items, which are morphologically negative, and are licensed syntactically rather than semantically. This claim is based on the observation that n-words are licensed under clausemate negation and not in constituent questions or conditionals. I will introduce polar questions to this picture and show that  a well-defined class of polar questions is also an n-word licensing context. I will further show that this class is intimately related to negation in a few other respects, from inducing polarity alternatives to morphosyntactic behavior. The question is how the implicit negation in these “bipolar” polar questions is encoded in the grammar. Do they rely on a syntactic head and consequently engage in syntactic concord, or should we rather re-analyze concord under negation as a special case of semantic licensing?

Speaker: Ora Matushansky (CNRS)

Title: Matching locatives with places

Date: Thursday 9 March

Venue: Lipsius 30

Time: 15.15-16.30 hrs

Abstract

An extremely robust but previously unnoticed cross-linguistic generalization is that the distribution of locative cases is frequently restricted to a specific subset of DPs: demonstratives, some classes of toponyms and a closed set of common nouns. I will provide examples from a number of languages of how exactly this restriction can play out and argue for the availability of locus denotation for some nouns, explaining this phenomenon.

Speaker: Güliz Güneş (Leiden University) & Aslı Göksel (Boğaziçi University)

Title: Phonology of Agreement: optional double agreement and optional resizing of phonological words in Turkish

Date: Thursday 23 February

Time: 15.15-16.30

Venue: Lipsius 30

Abstract:

THE ISSUE: Stress in Turkish is typically word-final (1a-b, where prominence is denoted by capital letters). However, certain morphemes trigger stress on the morphemes that immediately precede them (e.g. the copula in 1c). Given the exceptional stress patterns that they create, these so-called pre-stressing morphemes have generated much recent debate.

1. a. [kitAp] b. [gör-dü-lEr] c. [gör-ecEk-i-di-ler]

        book              see-PST-3PL      see-FUT –COP–PST-3PL

‘book’ ‘They saw (it).’ ‘They were going to see (it).’

2. a. [(gör-ecEk)M/ω -i-di-ler] b. [(gör-dü-lEr)M/ω] c. [(kitAp)M/ω]

The prosodic grouping in (2) hints towards an apparent mismatch of domains, in which a prosodic word may target an entire “word”, i.e. root+its affixes (in 2b and c, the verbal complex and the vowel harmonic word) or only a part of it (2a). In this paper we argue that such morphemes are not problematic after all. We claim that the phenomenon in (1c) arises not from a mismatch of domains, but from a mismatch in the timing and direction of the formation of morphosyntactic and prosodic constituents.

SET I – optional resizing of ωs: Any account that assumes the raising of the participle to a position where agreement resides (cf. Shwayder 2015 and the references in there) cannot foresee the variable prosodic grouping given below. This is because, if each M-word corresponds to an ω, and if an M-word is derived via raising, then any M-word that contains the agreement morpheme must invariably correspond to a single ω. This is not borne out – compare (3a) and (3b).

3. A: Had the visitors seen Bill?

a. B: [((gör-mÜş)ω-NP (-ler-Ø-di)ω)φ]ι-F

            see-PERF– 3PL–COP–PST

           Lit: ‘(They) had seen.’

a. B: [((gör-müş -lEr)ω-NP (-Ø-di)ω)φ]ι-F

            see-PERF-3PL –COP–PST

         Lit: ‘(They) had seen.’

Dissimilar to the previous accounts, we claim that there is no v-to-Asp or v-to-T raising in Turkish. As such, the M-words that host both the verb root and T or Asp are derived via the lowering of T and Asp to v. In the talk, we will show how this twist naturally accounts for the optional resizing of ωs given in (3).

SET II – optional double agreement: Although the agreement morpheme is commonly exponed on the right edge of the verbal complex (as in 1b and c), certain agreement markers may also be realized on the right edge of the verb participle (i.e. in the middle of the vowel harmonic word) (as in 3). Among the two agreement paradigms in Turkish, (i.e. K-paradigm and Z-paradigm), only the exponents of the K-paradigm and the third person plural form of the Z-paradigm seem to be allowed in the verb-medial position (Sezer 1998, Good & Yu 2005). What is more, prosodic grouping indicates that variable ω-formation is possible only with those agreement morphemes that are allowed in medial position.

We present novel data to show that the agreement markers that are allowed in the verb-medial position may also be realized simultaneously in two places, i.e. both verb-finally and verb-medially.

4. gör-müş –ler-Ø-di-ler.

    see-PERF–3PL–COP–PST–3PL

    Lit: ‘(They) had seen.’

In the second half of the talk we will discuss if there is a correlation between the medial (double) vs. final realization of agreement and prosodic resizing. We conclude that these are independent phenomena and however agreement is derived, a strategy such as lowering is necessary to predict variable parses given in (3).

We will also discuss other phonological accounts for double agreement from other languages such as Spanish (Harris & Halle 2005) and Basque (Arregi & Nevins 2012), and conclude that the double realization of agreement in Turkish cannot be accounted for by simply appealing to the phonological reduplication.

SELECTED REFERENCES: Good, J. & Yu, A. 2005. Morphosyntax of two Turkish subject pronominal paradigms. In Clitic and Affix Combinations, eds. L. Heggie, & F. Ordòñez: Amsterdam/ Philadelphia, John Benjamins B. V., 315-342. ■ Newell, H. 2008. Aspects of the morphology and phonology of phases. PhD dissertation, McGillU. ■ Sağ, Y. 2013. Copula in Turkish. In WAFL 8. Ed. U. Özge, 293–299. MIT WPL. ■ Shwayder, K. 2015. Words and subwords: phonology in a piece-based syntactic morphology. PhD dissertation, UPenn.

Speaker: Seid Tvica (UvA)
Title: The Rich Agreement Hypothesis beyond Indo-European
Date: Thursday 9 February
Venue: Lipsius 30
Time: 15.15-16.30 hrs

It is well-established in the literature that many Germanic and Romance languages differ in the placement of adverbs, appearing either before or after the finite verb. This typological distinction is standardly accounted for via v-to-I⁰ movement, arguably triggered by the subject agreement features that are assumed to be located at I⁰ (cf. Roberts 1985; Kosmeijer 1986; Rohrbacher 1994; Vikner 1995; Bobaljik and Thráinsson 1998; Koeneman and Zeijlstra 2014, among many others). The observed correlation between the properties of agreement morphology and verb movement gave rise to the so-called “Rich Agreement Hypothesis” (RAH) which states that in controlled environments the finite verb moves to a vP-external position if and only if the agreement morphology is rich (cf. Koeneman and Zeijlstra 2014).  In this talk, I present the results of a typological investigation of the RAH, showing that the hypothesis holds in all studied  languages outside Indo-European that exhibit environments in which the presence (or absence) of verb movement can be detected. Although a number of languages prima facie appear to challenge the RAH, it turns out that these “counter-examples” arise due to either morphological triggers of verb movement that are not related to agreement, or other types of displacements, such as vP- and vP-remnant movement.

Speaker: Metin Bağrıaçık (Ghent University)
Title: Relative Clauses in Asia Minor Greek
Date: Thursday 8 December
Venue: Matthias de Vrieshof 4/012
Time: 15.15-16.30 hrs

In this talk I will provide a synchronic and a diachronic account of the micro-variation in headed relative clause (HRC) formation strategies within the Asia Minor Greek dialects (dialects which were spoken in the regions known as Cappadocia, Pharasa and Pontus in modern-day Turkey until 1923, but are now spoken in various locations in Greece). In all the three dialects, HRCs are finite structures. In Cappadocian and (certain sub-varieties of Pontic) HRCs are obligatorily prenominal and are introduced by morphemes identical with the definite neuter singular/plural articles (‘to/ta’– dubbed as relative articles) which agree with the number feature on the head. In Pharasiot today, HRCs, which are –similar to Cappadocian and Pontic – finite, can nevertheless be both prenominal and postnominal without any semantic or pragmatic difference between the two, and the latter has been available only for the last four decades. Moreover, in this variety, both types of HRCs are introduced by a unique morpheme near identical in form to the relative articles of the other two dialects, ‘tu’.

I will first reveal that synchronically there is nothing as ‘relative articles’ in any of the dialects, but HRCs are simply subject to the phenomenon of ‘Obligatory Definiteness Spread’ according to which all (strictly) prenominal adjectival, numeral and relative modifiers have to be marked by overt neuter articles if the head they modify is definite. This is corroborated by the fact that there is a definiteness restriction on the head in HRCs in Cappadocian and Pontic. In Pharasiot, however, the head can be both definite or indefinite, which suggests that ‘tu’ is not an article, nor it is a relative article. Then, I will trace the origins of HRCs in Cappadocian/Pontic and Pharasiot in Early Medieval Greek and reveal that HRCs in Cappadocian/Pontic are successors of an already existing prenominal HRC structure with ‘relative articles’ but are by no means identical to this structure today. In Pharasiot, on the other hand, the ‘tu’ morpheme introducing HRCs is historically bimorphemic, composed of an external determiner from the neuter set (t-) and an invariant complementizer (u) that was employed to introduce HRCs. The morphological merger of these two yielded the prenominal HRCs in Pharasiot, which are by hypothesis of matching type. Tests applied on modern-day Pharasiot HRCs support the idea that these prenominal HRCs are indeed matching HRCs. On the other hand, the newly emerged postnominal HRCs, I argue, emerged due to pattern borrowing from Modern Greek and to identification of ‘tu’ with the Modern Greek invariant relativizer ‘pu’. This is witnessed by the fact that postnominal HRCs in Pharasiot are raising HRCs, precisely as Modern Greek HRCs with the invariant complementizer ‘pu’ (and without resumptives). In effect then, in modern-day Pharasiot, prenominal and postnominal HRCs differ not only in terms of their linear word order but also in terms of their internal structure.

All in all, the study casts serious doubts on the long-standing assumptions that (i) HRCs in the dialects are prenominal (solely) due to the Turkish influence exerted on them fr almost a millennium, and that (ii) Pharasiot is a sub-variety of Pontic or Cappadocian.

Speaker: Lieven Danckaert (CNRS/Université de Lille 3)
Title: The syntax-prosody mapping in a dead language, the case of Late Latin BE-periphrases
Date: Thursday 10 November
Venue: Matthias de Vrieshof 4/012
Time: 15.15-16.30 hrs

In this talk I will be concerned with certain aspects of Latin word order and in particular the alternation between the word order patterns ‘lexical verb – auxiliary’ and ‘auxiliary – lexical verb’. As is well known, in the present day Romance languages only the latter pattern is available. My starting point is the observation that in the particular case of periphrastic expressions with a BE-auxiliary and a past participle, Late Latin has a remarkably strong and uniform preference for the head-final order ‘PaPa – Aux’. In contrast, in the case of combinations of a modal verb and a dependent infinitive, we do in fact observe the expected rise of the head-initial pattern. In order to explain the unexpected behavior of BE-periphrases, I will develop an account which characterizes the Late Latin BE-auxiliary as a prosodically weak element, which is subject to a specific PF-condition which limits its positional freedom. Crucial evidence to support this analysis comes from prose rhythm, as well as from novel corpus data on the interaction between verb placement and negation.

Speaker: Olaf Koeneman (Radboud University)

Joint work with: Hedde Zeijlstra (University of Göttingen)

Title: The Rich Agreement Hypothesis: a language universal?

Date: Thursday 27 October

Venue: Lipsius 235C

Time: 15.15-16.30 hrs

Abstract

After its initial success, the hypothesis that verb movement to an inflectional position in the clausal structure, I(nfl), correlates with richness of morphology in some sense (Kosmeijer 1986, Rohrbacher 1994, Koeneman 2000, among others), has been criticized to such an extent that it almost became the laughing stock of generative syntax. In this talk, however, we will defend this Rich Agreement Hypothesis in its strongest, bidirectional formulation (V to I movement takes place if and only if agreement is rich) and argue that all prominent counterexamples have been misanalyzed. This puts the hypothesis in the position of a potential universal, and entails that there is something to be explained (which we will do) and something to be explored for the languages that are not part of the Indo-European language group (which Seid Tvica (University of Amsterdam) has done). Apart from discussing these consequences, we will also address some recent criticisms directed against our work.

Speaker: Sjef Barbiers (Leiden University)

Title: Restructuring Bridges

Date: Thursday 20 October

Venue: Van Eyckhof 2/005

Time: 15.15-16.30 hrs

Abstract

In the sentences in (1a,b) the MoodSpeechAct adverb eerlijk gezegd/honestly is surfacing in the embedded clause but must be interpreted in the matrix clause.

(1)         Syntax-semantics mismatch type 1

1. Ik denk dat ze eerlijk gezegd voor Rooney kiezen.                                   Dutch
2. I think that they honestly will go for Rooney.                                            English

Both: ‘I honestly think they will go for Rooney.’

The reverse type of syntax-semantics mismatch is found in (2). Here the adverb ook ‘also’ is surfacing in the matrix clause but can be interpreted in the embedded clause.

(2)         Syntax-semantics mismatch type 2

Ze hebben gebeld maar… ‘They have called, but…’                                          Dutch

Ik denk ook dat ze komen.

I think also that they come

‘I think that they will also come.’

The two types of mismatches do not have the properties of speech errors or parentheticals. They are both restricted to bridge verbs. The type of adverbs that can occur in mismatch type 1 depends on the bridge verb, e.g., the bridge verb THINK only allows for MoodSpeechAct adverbs, while the bridge verb WANT only allows for ModVolition adverbs. I argue that these properties follow if we extend Cinque’s (2001) analysis of restructuring to finite complement clauses, such that the “matrix” bridge verb originates in the embedded clause and moves to the position of the matrix verb. It is this movement that distinguishes bridge verbs from other CP-selecting verbs. Only adverbs located in the specifier of the base position of the “matrix” verb in the embedded clause must be interpreted in the matrix clause. Mismatch type 2 is restricted to adverbs that are generated in or can reach a position immediately below the base position of the “matrix” verb in the embedded clause. Movement of the “matrix” bridge verb into the matrix clause makes it possible for them to move into the matrix clause as well. A consequence of this analysis is that the syntactic structure of finite CP complements of bridge verbs depends on the type of bridge verb involved. If time permits, I will also discuss other consequences, such as the analysis of long Wh-extraction, Neg-raising and complementizers.