ComSyn Lectures

The next Comparative Syntax Meeting will take place on Thursday 17 October.

Speaker: Dafina Ratiu (University of Nantes)
Topic: Sharing at the syntax-semantics interface
Time: 13:15 – 15:00
Venue: Huizinga/023

Abstract
I discuss coordinated questions in Romanian, where two selected WHs (a subject & an object) appear coordinated in clause-initial position and I argue for a multidominant analysis I further address the question of the interpretation of multidominant structures. I propose that the shared material is interpreted in both conjuncts simultaneously ultidominance changes what counts as a constituent  I argue that, once we allow two rooted constituents and on the assumption that the interpretation feeds on the syntax then we need a step in the semantic interpretation which interprets a two rooted constituent as such.

Speaker: James Griffiths (Rijksuniversiteit Groningen)
Title: Comment clauses and Turkish ki
Date: Thursday 10th October
Venue: Eyckhof 1/003C
Time: 13:15-15

Abstract
Both hypotactic (1b) and paratactic accounts (1c) of how comment clauses such as I reckon and their hosts in (1a) are related have been advanced in the previous literature.

(1) a) John will be late, I reckon.
b) [[John will be late]1 [I reckon t1]]
c) [John will be late]i [I reckon Øi]

In the first half of this talk, I review some evidence in favour of adopting (1c) over (1b) for English comment clauses. In particular I introduce the par-Merge approach to parataxis (De Vries 2007), which utilises a novel functional head called Par0. Par0 is the locus for Potts’ (2005) COMMA feature and for parenthetical coordinators such as those in (2).

(2) a) John has – (and) it won’t be the last time – stolen a car.
b) John likes music, (but) especially Jazz.

I also show that, while the par-Merge approach offers conceptual advantages and a locus for coordinators, prime facie its extension to comment clauses is not well-motivated from an empirical perspective.
In the second half of this talk, I examine Turkish ki-clauses like (3), which are typically understood as finite subordinate clauses (Underhill 1976, Erguvanlı 1981, Göksel & Kerslake 2005, a.o.): an unexpected structure in a head-final language that otherwise nominalises non-root clauses.

(3) Hasan san-ıyor [ki Ali Bey Mine-yi taciz et-ti].
Hasan believe-PROG ki Ali Mr. Mine-ACC harassment make-PST
‘Hasan believes that Mr. Ali harassed Mine.’

I argue that the purported matrix clause in (3) equates with an English comment clause, and that ki is not a complementizer (as is usually assumed), but a morphological spell-out of Par0. If this is true, Turkish ki-clauses provide indirect evidence from an unrelated language for the adoption of the par-Merge approach to English comment clauses.

References
Erguvanlı, E. 1981. A Case of Syntactic Change: ki constructions in Turkish. Boğaziçi Üniversitesi, Beşeri Bilimler Dergisi Cilt 8:111-140.
Göksel, A. & Kerslake, C. 2005. Turkish: A comprehensive grammar. London: Routledge.
Kornfilt, J. 1997. Turkish. Routledge.
Potts, C. 2005. The Logic of Conventional Implicatures. OUP
Vries, M. de. 2007. Invisible Constituents? Parentheses as B-Merged Adverbial Phrases. In Parentheticals, ed. by Dehé, N. & Kavalova, Y,. Amsterdam: John Benjamins. 203-234.

Speaker: Jean Crawford (University of Connecticut)
Title: Verbal passives in child English: Evidence from judgments of purpose phrases
Venue: Huizinga/o25
Time: 15:15-16:30

Abstract
Most maturational accounts for passive acquisition claim that the passives seen in early child speech are not adult-like verbal passives, but rather an adjectival construction with a simpler syntax (Borer and Wexler 1987, Babyonyshev et al. 2001). The Universal Phase Requirement (UPR, Wexler 2004) assumes children use resultative adjectival passive syntax (Embick 2004). The Argument Intervention Hypothesis (AIH, Orfitelli 2012) must assume that children’s good performance on short and long actional passives is due to a syntax that does not have an intervening agent argument. Alternatively, Snyder and Hyams (2008) argue that children’s syntax for verbal passives is intact, but passive movement will violate relativized minimality unless the context adds discourse features to one of the arguments to distinguish the chains (Rizzi 2004).
Verbal and adjectival passives in English can be disambiguated with purpose phrases (PPs). Verbal passives contain a syntactically represented implicit argument (IMP), which can license a PP (Roberts 1987). PPs are allowed with actives and verbal passives, but not with middle/inchoative constructions.

(1) a. John is breaking the candy bar to share with friends
b. The candy bar is being broken IMP PRO to share with friends
c. *The candy bar is breaking to share with friends

PPs are also not acceptable with adjectival passives because they do not have IMPs to control PRO:

(2) *The candy bar is unbroken to share with friends

If young children’s passives are verbal, they should judge passives with PPs like (1b) as acceptable, just like they do (1a). If children’s passives are adjectival and do not contain an intervening IMP, they should judge (1b) to be as unacceptable as (1c). As PP acceptability among the constructions is based on grammaticality judgment (GJ) data from adults, it seems appropriate to evaluate children’s knowledge with a similar judgment task.
Twenty-one 4-6-year-olds participated in a targeted GJ task (Stromswold 1990, McDaniel and Cairns 1996, Hiramatsu 2000). After a training and pretest which focused on judging active and inchoative forms, children provided judgments for 5 verbs (bake, break, grow, light, sink) in 4 different constructions (active progressive, passive progressive, inchoative progressive, inchoative present). Each item was presented with a story emphasizing the object. Passives were the critical items. Paired t-tests showed children judged passives differently from both types of inchoatives (progressive inchoative: t(1,20)=3.25, p=.004, present inchoatives: t(1,20)=5.59, p<.001). A repeated measures ANOVA on all 4 constructions reveals a main effect of verb type (Greenhouse-Geisser F(2.202,43.663)=33.660, p<.001). Post-hoc pairwise comparisons (with Bonferroni correction) showed passives varied significantly from both types of inchoatives. The inchoatives did not vary significantly from one another. Like adult controls, passives also varied significantly from actives.
The results show that children use verbal passive syntax to comprehend passives, providing evidence against the UPR and AIH. The results provide preliminary evidence for Snyder and Hyams’ proposal, though this account faces other challenges (Crawford 2012). Following Grillo (2008), I propose children’s difficulties with passives may stem from constructing the complex event structure required for passives of activity and stative predicates.

Speaker: Ekaterina Chernova  (Universitat de Girona)
Title: On the typology of (multiple) wh-fronting from a Q-based perspective
Venue: Lipsius/235b
Time: 15:15-16:30

Abstract
In this talk, I will try to account on the nature of multiple wh-fronting (MWF) in Slavic languages, where all wh-items overtly move to the edge of a clause. However, although superficially very similar, MWF languages behave differently in numerous ways regarding wh-movement: while Bulgarian shows Superiority effects and forces the intervening material to be placed after the whole set of fronted wh-items, Russian behaves just the other way round. Thus, Russian has been claimed to be in its core a wh-in-situ language where wh-fronting is [focus]-driven, while Bulgarian has been argued to exhibit true [wh]-driven movement (see Stepanov 1998; Bošković 2002). However, I will try to show that there is an alternative way to explain this complex set of data withoutappealing to the weak focus- vs. wh-movement distinction. Following Cable (2010), I argue that what has been analyzed as wh-/focus-movement is in fact a secondary effect of Q-movement, when a Q-particle merged with a wh-item first takes it as its complement projecting a QP; then attraction of Q’s features into CP triggers movement of the whole QP, no wh-feature-percolation being necessary. I will propose some minimal, yet significant, additions to the original Cable’s Q-theory arguing that it cannot capture the MWF phenomenon in its current formulation.

In the second part of the talk, I will try to present a more general picture of Q-movement patterns observed in multiple wh-questions across languages. I will briefly consider Spanish, which apparently contradicts Cable’s generalization on a correlation between Superiority and Intervention effects, since it can violate both, and propose a syntactic account. Finally, a tentative parametrical hierarchy of crosslinguistic variation with respect to Q-movement will be offered.

Title: Complementizer-trace effects: a crosslinguistic perspective
Speakers: E. Boef  (ZAS Berlin) & Irene Franco (LUCL Leiden)
Venue: Lipsius/235b
Time: 15:15-16:30

Abstract
In this talk we present ongoing research on the well-known complementizer-trace (COMP-t) effect in the lower clause of long-distance A-bar dependencies (specifically relative clauses and wh-questions). The COMP-t effect, by which a finite declarative complementizer cannot be directly followed by a trace is illustrated in (1) for standard English.

(1)       a.         I know the man that you said  (*that) t came here.
b.       I know the man that you said (that) Mary will meet t tomorrow.

The COMP-t effect is certainly not a universal phenomenon, and its absence has traditionally been related to the availability of null subjects in a language. However, such a generalization does not hold universally, as can be seen in Icelandic in (2).

(2)      a.         Ég þekki manninn sem þú   sagðir %(að) kom   hingað.
I    know man.the  sem you said that came here
b.         Ég hata manninn sem þú sagðir %(að) María ætlar að hitta t á  morgun.
I  hate man.the sem you said that Maria is.going to meet  tomorrow

We provide a cross-linguistic typology of the COMP-t effect starting from Germanic, and extending to Romance. We show that besides the null subject parameter, other morphosyntactic properties play a role as well, such as the presence of V-to-C movement in embedded extraction contexts and D-morphology on the declarative complementizer.

We propose that the lowest C-head in the left periphery (Fin⁰) encodes a D-feature that requires checking in order to make the subject interpretable: in Germanic this checking requires move or merge to FinP, whereas in (some) Romance languages D can be checked by the subject in IP.

In extraction contexts (where COMP-t effects emerge), there are several ways to check D on FinP: (i) expletive/resumptive pronoun insertion;  (ii) V-to-Fin movement; (iii) D-morphology on the complementizer. If none of these strategies are available in a language, the complementizer needs to be dropped in order for the D-feature to be checked higher up the A-bar chain. Put differently, the complementizer in this case acts as a sort of intervener.

Speaker: Sjef Barbiers (Meertens Instituut and University of Utrecht)
Title: Landing sites and stranding sites
Venue: Lipsius/235B
Time: 15:00 – 16:30

There is a growing body of evidence for Chomsky’s claim (Chomsky 1986 and subsequent work) that vP is an intermediate landing site for long distance (LD) movement, cf. Barbiers (2002) for Dutch, Rackowski and Richards (2005) for Tagalog, Den Dikken (2009) for Hungarian and Koopman (2010) for West Ulster English and Dutch. Traditionally, embedded SpecCP is taken to be an intermediate landing site as well. The first claim of this talk, extending Barbiers (2002), is that these two types of intermediate landing sites behave categorically distinct on a series of stranding tests (P-stranding, Dutch floating quantifier  zoal ‘so all’, Dutch floating quantifier allemaal ‘all’, focus particle maar ‘only’, wat voor ‘what for’ split and remnant indefinite DP). Stranding is never possible in embedded SpecCP and always possible in vP, both embedded and matrix. With doubling we find the opposite pattern: doubling is possible in embedded SpecCP but never in vP (cf. Barbiers, Koeneman and Lekakou 2009). I argue that (sub-)extraction from embedded SpecCP is impossible because CP is dominated by the phase vP. Subextraction from this position can only give rise to a converging derivation if the offending copy in SpecCP is spelled out, yielding doubling (cf. Boef 2013 and van Craenenbroeck and van Koppen 2008). The second claim of this talk is that the picture according to which vP is the landing site for an intermediate movement step is too simple. Using the hierarchy of projections as proposed in Cinque (1999) I show that at least 4 different intermediate landing sites have to be distinguished. Which one is chosen depends on the type of constituent that moves through it, as can be made visible with stranded material.

Title: Configurationality and Word Order in Latin and Romance
Speaker: Adam Ledgeway (University of Cambridge)
Time: 15:00 – 16:30
Venue: Lipsius 235/b

Among Romance linguists of all theoretical persuasions, there is general recognition
that, in the passage from Latin to Romance, the morphosyntax of the emerging
languages underwent significant changes in three fundamental areas of the grammar
involving the nominal group, the verbal group, and the sentence. The impact of such
changes is most immediately observable in the emergence of a series of functional
categories (including determiners, auxiliaries, and complementizers) and in the
gradual rigidification of word order in these same groups. Now, while the specific
details of the complex morphosyntactic changes affecting the three key areas of the
grammar are relatively well known, scholars are still very much divided as to their
correct interpretation, and how they are to be integrated within the overall typological
changes witnessed in the passage from Latin to Romance. Traditionally, the principal
typological difference between Latin and Romance has been taken to involve a
distinction between morphology and syntax: while Latin predominantly makes
recourse to synthetic structures (with concomitant so-called free word order), the
morphologically poorer Romance varieties make greater use of analytic structures
(with concomitant fixed word order). In this talk I shall reconsider this traditional
theme of Romance linguistics, showing that the predominant analytic patterns of
Romance are nothing more than the partial reflex of a more deep-rooted structural
change.
According to one view, this change involves a move from non-configurationality to
full configurationality: whereas in Latin grammatical relations are encoded by the
forms of words themselves through case and agreement morphology, so-called
lexocentricity (Bresnan 2001: 109-112), in Romance grammatical relations are
encoded through the syntactic context of individual words organized into distinct
hierarchical phrase structure configurations. Indeed, as Vincent (1998: 423f.)
observes, Latin presents all of Hale’s (1983) classic tests for non-configurationality
originally established on the evidence of Warlpiri (see also Ledgeway 2011: §3.4). On
this view, the emergence of functional categories in Romance can be seen as a
concomitant of the emergence of hierarchical constituent structure in the nominal,
verbal, and sentential domains which makes available a position for functional
elements such as determiners (DP), auxiliaries (IP/TP) and complementizers (CP),
thereby reflecting the traditional intuition popularized within the synthesis-analysis
approach which highlights the emergence in Romance of articles and clitics,
auxiliaries, and a whole host of finite and non-finite complementizers, all generally
absent from Latin.
Despite the merits of this view of the Latin-Romance development in terms of the
rise of (full) configurationality, I shall develop an alternative approach to the changes
in word order and argument realization from Latin to Romance which assumes the
presence of both configurational and functional structure already in Latin. On this
view, the unmistakable differences between Latin and Romance, most notably
observable in the replacement of an essentially pragmatically-determined word order
with an increasingly grammatically-determined word order and the concomitant
emergence of functional categories, can now be explained by formal changes in the
directionality parameter and the differential role of functional structure in the two
varieties. The gradual rigidification of word order according to grammatical principles
in the passage from Latin to Romance can be explained in terms of a progressive
reversal of the directionality parameter: assuming the ordering of heads and
complements in the development from (Indo-European/) archaic Latin to Romance to
have undergone a shift from one harmonic principle of linear organization to another
(viz. head-last ⇒ head-first), the greater freedom of word order traditionally
recognized for classical Latin can now been seen as a result of its occupying an
artificially sustained intermediate position in this change, resulting in mixed
(dis)harmonic linearizations. Besides this fluctuation at the syntactic level between a
conservative head-final and an innovative head-initial structural organization,
pragmatics is also widely recognized to play a significant role in determining Latin
word order. This aspect of Latin sentential organization, largely absent in Romance,
can be captured by assuming the greater accessibility of topic- and focus-fronting to
left-peripheral positions situated in the left edge of individual functional projections.
In Romance functional structure is readily exploited and made visible through the
lexicalization of head positions with functional categories such as determiners,
auxiliaries and complementizers, as well as through operations such as N(oun)- or
V(erb)-raising to these same head positions. By contrast, Latin lacks such functional
categories and N-/V-raising, but displays ubiquitous evidence for the presence of
functional structure through its extensive exploitation of topic- and focus-fronting to
the left edge of these same functional projections.
Adopting this view, the perceived non-configurationality of Latin can be broken
down into two main ingredients: i) grammatically-free word order resulting from an
ongoing change in the head directionality parameter (ultimately interpreted as the
progressive loss of Complement-to-Specifier roll-up movement), which a priori
allows dependents/complements to occur on either side of their head; and ii)
pragmatically-driven word order, often producing discontinuous structures, resulting
from the greater accessibility of topic- and focus-fronting to positions situated in the
left edge of individual functional projections. Interpreted in this manner, the apparent
emergence of configurationality in Romance is to be understood as the surface effect
of the rigidification of the directionality parameter and the restricted accessibility of
edge-fronting to left-peripheral positions within the functional structure. In short, it is
these formal changes in the directionality parameter and the differential role of
functional structure in the two varieties which conspire to give the superficial
impression

The next ComSyn meeting will take place on Tuesday 12 February.

Speaker: Petra Sleeman ( University of Amsterdam)
Title: Local and non-local gender agreement in French (joint work with Tabea Ihsane, University of Geneva)
Time: 15:15- 16:30
Venue: Lipsius 235b

Abstract

Whereas in English there are no gender distinctions for the noun, in Romance languages masculine and feminine nouns can be distinguished, both for animate and inanimate nouns. Although for inanimate nouns gender is always completely arbitrary, for animate nouns morphological gender does not always correspond to natural semantic gender, as illustrated by French (la sentinelle (f.) ‘the sentinel’ and le médecin (m.) ‘the doctor’, le professeur (m.) ‘the professor’, le mannequin ‘the model’: all refer to men and women).

            To account for this dual behavior of nouns (arbitrary or natural gender), it has been proposed in recent literature (Kramer 2009; Atkinson 2012) that gender is expressed in two positions within the DP, and not in one (either the root or n, as in Alexiadou 2004 or in Lowenstamm 2008, resp.): as an uninterpretable feature on the root (accounting for arbitrary gender: le magasin (m.) ‘the shop’; la sentinelle (f.) ‘the sentinel’) or as an interpretable feature on the head of the functional projection nP (accounting for natural/semantic gender: un homme (m.) ‘a man’; une chatte (f.) ‘a (female) cat’; une enfant (f.) ‘a (female) child’; un enfant (m.) ‘a (male) child’). Determiners and adjectives agree with the ±masc. feature on n (une chatte (f.) blanche ‘a white cat’), or with the gender feature that is specified on the root (un vieux magasin (m.) ‘an old shop’, la malheureuse sentinelle (f.) ‘the poor sentinel’). In the case of the non-specified noun un enfant ‘a (male or female) child’ / un chat ‘a (male or female) cat’, the determiner and the adjective agree with the root, which gets default masculine gender.

            This analysis seems to account quite nicely for the local agreement between D, adjectives, and the noun in D-A-N or D-N-A configurations. In this paper we put this analysis, however, under scrutiny by investigating whether it can also account for less local agreement configurations in French. We argue that if agreement takes place with an element outside the strict DP phase, gender on the root can (1), must (2) or cannot (3) be overridden by a supplementary interpretable gender feature on n. We base our account on valuation and the interpretable/uninterpretable character of the gender feature (Pesetsky & Torrego 2007).

The next ComSyn meeting will take place on Tuesday 4 December.

Speaker: Erik Schoorlemmer (LUCL)
Title: The Prohibition that never was. Extraction, blocking and genitive noun complements in Serbo-Croatian.
Time: 15:15-16:30
Venue: Lipsius 235c

Abstract

In Serbo-Croatian, sentences in which a genitive form of the Serbo-Croatian counterpart of who is extracted out of a noun complement position are ungrammatical (Zlatić 1997, Bašić 2004, Bošković 2011). This has been taken as evidence that there is a general prohibition on the extraction of noun complements in Serbo-Croatian (see among others Bošković 2011). In this talk, I, however, claim that such a prohibition doesn’t exist. Instead, I will show that the ungrammaticality of extracting genitive who is due to a more general effect: the use of a genitive form is blocked if a prenominal possessive form can be used. I will propose a tentative account of this blocking in terms of head movement.

The next ComSyn meeting will take place on Tuesday 20 November.

Speaker: Luisa Meroni (Universiteit Utrecht)
Title: Specific indefinite objects in bilingual acquisition.
Time: 15:15-16:30
Venue:  Lipsius 235c

Abstract

Crosslinguistic influence in simultaneous bilinguals (2L1) has been claimed to lead to delay and acceleration (Paradis & Genesee 1996). Whilst various studies have demonstrated the existence of delay, the evidence for acceleration is rather limited (cf. Meisel 2007). Furthermore, most studies in this area focused on the acquisition of morphosyntax and syntax-pragmatics; to date, few examine the area of syntax-semantics (but cf. Serratrice et al. 2009, Unsworth 2012). This paper seeks to fill these gaps by investigating whether crosslinguistic influence in the form of acceleration takes place in the interpretation of specific indefinites in sentences with negation by 2L1 Dutch-Italian children.

Previous research has shown that Dutch-speaking 4- to 6-year-old children pass through a stage, where, unlike adults, they treat the scrambled sentence in (1) as ambiguous between a specific (1a) and a non-specific (1b) interpretation (Unsworth et al., 2008).

(1) Ian heeft een kaarsje niet uitgeblazen
a. There is a candle Ian did not blow out (a>not)
b. Ian did not blow out any candle (not>a)

In a recent study, Unsworth (2012) shows that English-Dutch 2L1 children pass through a similar stage and they furthermore restrict (1) to its target specific interpretation within the same timeframe as monolinguals (L1). Here, we examine data from Italian-Dutch 2L1 children. Following Su (2001), we argue that, unlike English, indefinite objects in negative sentences in Italian (2) are unambiguously specific as a result of the same form (‘un/uno/una’) being used for both indefinite and numeral (i.e., ‘one’).

(2) Sandro non ha spento una candelina
‘Sandro did not blow out a candle’
a. There is a candle Sandro did not blow out (a>not)

We hypothesize that at the stage where child Dutch allows (1) to be associated with both specific and non-specific interpretations, and, in Italian, (2) is interpreted specifically, the availability of the specific interpretation in Italian will facilitate its acquisition in the Dutch of Italian-Dutch 2L1 children, i.e., we expect crosslinguistic influence in the form of acceleration.

We first show that (2) is indeed unambiguous by using a (picture) Truth Value Judgment task where sentences like (2) were evaluated in a situation in which Sandro blew out all the candles except one : L1 Italian (4;6-5;11; n=7) children almost always accepted the specific interpretation (92 % (32/35)). Subsequently, the same task was used in both languages with 2L1 Dutch-Italian (4;4-6;1; n=13) and in Dutch with age-matched L1 (n=13) children. The results, cf. Figure 1, show that: (i) like monolinguals, 2L1 children almost always accepted the specific interpretation in Italian (96 % (62/65); U(20)=35, p=.45); (ii) this was also the case for Dutch (acceptance rate: 97% (63/65)); and (iii) crucially, the 2L1 children accepted the specific reading significantly more often than the age-matched monolinguals (cf. 48% (34/65); U(26)=30, p=.001).

The data demonstrate crosslinguistic influence in the area of syntax-semantics and are in line with previous work on scope preferences in heritage learners (Lee et al. 2010; O’Grady et al. 2011); however, in contrast to this work, our data show acceleration rather than delay.