SEMANTICS IN ATHENS III

Workshop on semantics and the syntax-semantics interface

Saturday, December 21, 2019

National and Kapodistrian University of Athens

Athens University History Museum (Tholou 5)

Attendance is free. If you plan to join us for dinner (20:00, Το Καφενείο), please send us a brief message by Tuesday, December 17.

Organizers: Winfried Lechner (EKPA, Athens) & Giorgos Spathas (Leibniz-ZAS, Berlin)

Programm

09:00 – 09:10	Welcome
Session 1
09:10 – 09:50	Nikos Angelopoulos (KU Leuven)	Complementizers as Probes
09:50 – 10:30	Maria Barouni (University of Crete)	Absolute superlatives: evaluativity and beyond
10:30 – 11:10	Renos Georgiou (University of Patras)	Some problems with syntactic analyses of Greek CLLD
11:10 – 11:40	Coffee Break
Session 2
11:40 – 12:20	Maria Margarita Makri (University of York)	Types of phrasal comparatives: teasing apart syntax and semantics
12:20 – 13:00	Athanasios Michail Ramadanidis (University of Crete)	Some observations on the form and interpretation of spatial expressions
13:00 – 15:00	Lunch Break
Session 3
15:00 – 15:40	Stavroula Alaxandropoulou (Utrecht University)	wNPIs in non-downward entailing contexts: experimental investigation of possible contextual licensing
15:40 – 16:20	Stergios Chatzikyriakidis (University of Gothenburg)	To infer or not to infer: natural language inference, computational semantics, and theoretical linguistics
16:20 – 17:00	Viola Schmitt (University of Vienna)	Individuating individual concepts

Abstracts

Nikos Angelopoulos (KU Leuven)

Complementizers as Probes

In this talk, I present new data from Greek showing that complementizer, C, selection in declarative embedded clauses introduced by pu and oti is determined by interpretive properties of the matrix verb. Concretely, the new striking observation I bring to light is that pu can only introduce clauses after predicates which are stative. Given this, I propose a new analysis according to which Cs are merged in the matrix clause and they select the matrix verb instead of being selected by it (cf. Kayne 2000, 2005). Furthermore, under this analysis, Greek Cs are probes attracting their surface complement instead of merging directly with it. I show that this analysis fares better with the sensitivity Cs exhibit in regard to grammatical properties of the embedding verb as well as with many other interpretive and distributional properties of embedded clauses in Greek. On the other hand, I discuss that this set of facts are puzzling for the ‘standard’ analysis according to which Cs are merged in the left periphery of the embedded clause.

Maria Barouni (University of Crete)

Absolute superlatives: evaluativity and beyond

In this talk I will attempt to explore the evaluative component of absolute superlatives. I will start by arguing that absolute superlatives contain also a comparative morpheme, as opposed to the standard view (Pound, 1908; Bobaljik, 2012). I will show how a diachronic approach can shed light to our understanding of these elements. Moreover, I will explore the evaluative component of these elements, ‘more than the standard/ more than expected’’ in relation to other particles with evaluative component (i.e. only/even) (Greenberg 2019, von Fintel 2019).

Renos Georgiou (University of Patras)

Some problems with syntactic analyses of Greek CLLD

It is a standard assumption in generative grammar that Clitic Left Dislocation (CLLD) is sensitive to strong islands (Cinque 1990, Iatridou 1995, Anagnostopoulou 1994). The present study shows that this is not necessarily the case, as the left-dislocated phrase may cross a relative clause or an adjunct (island). In this context we also examine the ‘reconstruction’ properties of Greek CLLD. It is shown that the reconstruction effects typically found with CLLD are affected by the intervention of islands. The asymmetries that arise seem fairly problematic for the existing analyses. The data suggest a mixed-chain analysis (cf. Iatridou 1995, McCloskey 2002, Sportiche 2018) for CLLD dependencies that cross strong islands.

Maria Margarita Makri (University of York)

Types of phrasal comparatives: teasing apart syntax and semantics

In this talk I examine the syntactic properties of the standard phrases that are introduced by a prepositional standard marker. Careful examination of the types of phrasal standards allowed in Greek and Romance reveals that there is no one-to-one mapping between the phrasal vs. clausal distinction on the one hand and the choice between a two- and a three-place comparative operator on the other. French, Spanish, and Catalan phrasal standard markers select only for —degree— denoting Numeral Phrases, which do not necessitate the postulation of a distinct comparative operator. On the other hand, the Greek and Italian prepositional standard marker selects for DPs, NumPs, and AdjPs irrespective of their semantic type. I also review the implications of variation in the standard phrase for the status of phrasal standards in the comparative phrase and the semantic type of NPs.

Athanasios Michail Ramadanidis (University of Crete)

Some observations on the form and interpretation of spatial expressions

Expressing spatial relations and movement in space is a core domain of natural language. But how are they semantically modelled and syntactically encoded? The talk explores some aspects of the syntax and semantics of spatial constructions in Greek and other languages and argues that advances in vector space and path semantics provide important guidelines about the syntactic organization of spatial information.

Stavroula Alexandropoulou (Utrecht University; joint work with Lisa Bylinina and Rick Nouwen)

NPIs in non-downward entailing contexts: experimental investigation of possible contextual licensing

We investigate experimentally to what extent readers of sentences containing weak negative polarity items (wNPIs) in non-downward entailing (non-DE) environments make assumptions about the context. It has been argued, for instance, that the NPI ‘anything' in (1) is licensed in the scope of the non-monotonic quantifier exactly via contextual reasoning (relating to a higher expectation).

(1) Exactly two of the boxes contain anything.

Such reasoning is predicted by a number of theories of polarity licensing. Broadly, there exist three approaches to NPI licensing in non-DE environments: Giannakidou (2008), Crnič (2014), and Barker (2018). These theories share the idea that wNPIs are licit in non-DE environments only under specific contextual circumstances and crucially differ in the properties of these circumstances, giving rise to different predictions. Our experiments evaluate their predictions on wNPI licensing in environments created by the quantifiers at most, at least, exactly and between. On the basis of the data obtained from a series of offline experiments, we conclude that at least is a bad wNPI licensor, exactly is a good licensor on its own, and between is in need of contextual licensing. In our talk, we will argue that our overall findings are at odds with Giannakidou’s or Crnič’s predictions, but compatible with Barker’s theory. We will also consider the possibility that the attested discrepancy between the non-monotonic quantifiers exactly and between is due to a real interpretative difference.

Stergios Chatzikyriakidis (University of Gothenburg)

To infer or not to infer: natural language inference, computational semantics, and theoretical linguistics

In this talk, I will present an overview of the Natural Language Inference (NLI) task. I will go through the three main types of approaches that have been used in tackling the problem and discuss their advantages and weaknesses. A similar overview will be given for the datasets that have been used to train/evaluate these approaches. I will end with a discussion on the relevance of Theoretical Linguistics for NLI in the age of Neural Networks and argue for hybrid approaches to NLI.

Viola Schmitt (University of Vienna; joint work with Nina Haslinger)

Inviduating individual concepts

In this talk, we argue that the question of how we individuate inexistent objects (see eg Quine 1943) is a linguistically relevant one. This is already suggested by the `similarity'-problem raised Geach's 1967 Hob-Nob-sentences, however, we show that a number of other constructions also give rise to the question of when two inexistent objects count as distinct and that an intuitive notion of distinctness in such constructions matters for truth-vale judgements. We give a descriptive generalization aimed at constraining the relevant notion of distinctness.