December 4th, 2025, Sonia Ramotowska

Learnability of quantifiers and semantic universals

All natural languages share common properties called universals. In the domain of quantification, three semantic universals were discovered: monotonicity (convexity), quantity, and conservativity. Researchers have been trying to explain the origin of semantic universals for decades. In this study, we tested one of the proposed explanations, the learnability hypothesis. According to this hypothesis, quantifiers that satisfy universals are easier to learn and, therefore, more likely to be lexicalized in natural language. We tested the learnability hypothesis in a large-scale, online, between-subjects design experiment, in which participants learned a new quantifier, gleeb. Gleeb corresponded to one of the following quantifiers: monotone, convex, and quantitative at least 3 and at most 2, non-monotone and convex between 3 and 6, non-convex even number, non-quantitative the first 3, the last 3, conservative not all, and non-conservative not only. We found that all universals affected the speed of acquisition: participants learned conservative quantifiers faster than non-conservative quantifiers, quantitative quantifiers faster than non-quantitative quantifiers, and upward monotone quantifiers faster than non-monotone quantifiers (however, the lack of convexity did not affect the rate of learning more). In conclusion, our results provide evidence for the learnability hypothesis.

November 27th, 2025, Zhuoye Zhao

Tense as a perspective-sensitive item

Tense is commonly construed as a deictic category (Comrie, 1985). While much discussion has been devoted to its deixis, relatively little attention is paid to the nature of its deictic center. In joint work with Anastasia Tsilia (MIT), we propose that tense as a deictic category is comparable (though not identical) to indexicals, in that its interpretation depends on a parameter that has a global effect on an interpretation domain.

The proposal is primarily motivated by a ‘shift-together’ effect between tenses and the temporal adverbial then (in English as well as its cross-linguistic counterparts) that cuts across the previously-established typology of tense shift (Ogihara & Sharvit, 2012). The pattern is explained within a new theory of tense, according to which its interpretation is sensitive to a temporal perspective parameter, and tense shift happens by means of perspective shift. The new theory also demarcates the often-conflated notions of a shifted present and a deleted tense.

November 20th, 2025, Michael Goodale

Learning form and learning meaning: Modelling language acquisition using symbolically grounded corpora

Large language models show remarkable syntactic abilities with very little in-built structure after training on enormous corpora. They also miraculously follow instructions and appear to "understand" language to some extent. However, their semantic abilities are fundamentally an epiphenomenon born out of their training objective: next token prediction. This stands in sharp contrast to humans, whose acquisition and use of language is grounded in context where they are constantly conveying and interpreting meanings and intentions. Large language models' goal is to learn forms while humans are learning a mapping from forms to meanings.

In this talk, I will outline an alternative approach to modelling language acquisition which aims directly to learn interpretable meanings rather than solely predict grammatical strings. This approach uses "grounded corpora", datasets where sentences are paired with model-theoretic representations of a scene, rather than text alone or text paired with sensory data. A learning algorithm then aims to find symbolic grammars which compositionally assign meanings in a language of thought to strings. This algorithm will specifically be used to compare different theories of innate linguistic knowledge regarding the linking problem: the systematic mapping of thematic roles and syntactic position.

November 6th, 2025, Yair Lakretz

Developing and Evaluating Cognitively Plausible Neural Models for the Study of Language Processing in the Human Brain

Finding the neural basis of human language processing has been a major challenge in cognitive neuroscience for many years. Neural AI models now provide new opportunities to study the brain's neural encoding and mechanisms for language processing, as their internal activations and connections are fully observable. However, while originally inspired by the human brain, current neural models often diverge from their biological counterparts in terms of architecture, training methods, and data requirements. To be a useful scientific tool for cognitive neuroscience, we argue that we should develop more cognitively and biologically plausible models and avoid relying on commercial, off-the-shelf solutions. In this talk, I will discuss several concrete ways to make the models more human-like, and I will present an overview of ongoing studies in the lab where we are exploring these directions, in various domains, such as phonological and syntactic processing. I will conclude by showing examples from human data that serve as behavioral and neural benchmarks for model evaluation.

October 23rd, 2025, Isabelle Charnavel

A semantic treatment of indexical binding

Indexical binding in constructions illustrated in (1)-(2) is one of the crucial empirical arguments questioning the Kaplanian hypothesis that indexicals rigidly depend on context.

(1) Only I did my homework. (i.e., nobody else did their homework) (Heim 1991)

 (2) I am the only one around here who will admit that I could be wrong. (i.e., nobody else will admit that they could be wrong) (Partee 1989)

 The goal of the talk is to propose an analysis of such cases that maintains contextual dependency as a fundamental property of indexicals. The analysis, which is mainly based on French data, requires re-examining the role of binding, focus, agreement and presuppositions in these constructions.

October 9th, 2025, Paul Egré

Conditionals, Presuppositions, and Logic: Avoid Void

Trivalent semantics of indicative conditionals state that a sentence of the form "If A then C" is void or undefined when the antecedent A is false (de Finetti 1935, Quine 1950, Cooper 1968, Belnap 1970, 1973). It has been argued that a trivalent treatment of conditionals wrongly predicts conditionals to presuppose the truth of their antecedent, instead of the weaker presupposition that the antecedent should be compatible with the common ground (Stalnaker 1975, von Fintel 2011). We propose to derive this weaker presupposition from a specific version of Stalnaker's Bridge Principle between pragmatic and semantic presuppositions, which we call "Avoid Void". "Avoid Void" prohibits asserting a proposition that would be void throughout the context (see also Rossi and Santorio 2025, who call the principle "Nonvacuousness"). In the first part of this paper, we examine a number of predictions of this principle when used in combination with Cooper's semantics for the conditional (Cooper 1968; Egré, Rossi, Sprenger 2021, 2025), including some difficulties in relation to presupposition projection. In the second part of the paper, we use the same principle to filter out some problematic inferences that remain valid in Cooper's logic of indicative conditionals, but that appear to be uninformative in the same way. This includes the inference of Falsity of the Antecedent ("it rains, therefore if it doesn't rain, 2+2=4") and specific instances of Truth of the Consequent ("it rains, therefore if it doesn't rain, it rains"). A Strawsonian feature of the resulting logic is that from a set of contradictory premises, not only is it not the case that *anything* follows ("Ex contradictione quodlibet"), but it will typically happen that *nothing* follows ("Ex contradictione nihil"). (Joint work with Lorenzo Rossi and Jan Sprenger)

October 2nd, 2025, Zhuoye Zhao

Cross-domain functional particle: the case of Mandarin le

The semantics of the verbal-suffix le has long been a central puzzle in the study of the temporal system of Mandarin. In this talk, I argue that instead of a dedicated temporal morpheme, le should be construed as a cross-domain functional particle underspecified for the associated semantic domain.

September 25th, 2025, Carlo Geraci

Ohlala! Exclamatives vs. Exclamation in LSF: Some preliminary results

As part of a collective project to study Exclamatives in sign language we present today some very preliminary results on how exclamation is marked in LSF. We identify the main manual and non-manual strategies recurring in LSF, and we address whether these constructions have the signature of exclamative clauses, understood as an independent sentence type, distinct from questions and declaratives.

September 18th, 2025, Jeanne Lecavelier 

A new perspective on Remind-Me questions

Remind-Me questions (RMQs), i.e. questions like “What is your name again?”, have been defined by
Sauerland & Yatsushiro (2017) as questions asking the Addressee to repeat an information inside a discourse event.
While this analysis fits the use of presuppositional again as a trigger of RMQs, it does not extend well to other
languages that reuse temporal devices for RMQs, like Danish nu (or Finnish nyt, Estonian nüüd) ‘now’, or French déjà (or Slovenian že) ‘already/ever’.

This talk provides new evidence that the defining element of RMQs does not lie in the discourse event, but rather in a mismatch between the Common Ground and the Speaker’s knowledge. We propose a new characterisation of RMQs as non-canonical Qs conveying both past knowledge and current ignorance. Building on this, we offer two novel analyses for the Danish and the French cases, and show how they integrate into the cross-linguistic landscape of RMQ-triggers. (joint work with Alexander Wimmer, Universität Stuttgart)

April 17th, 2025, Lucas Champollion

A variably strict truthmaker semantics for modals and conditionals

This work in progress presents a novel approach to modals and conditionals by integrating two independently well-known frameworks: variably strict semantics and truthmaker semantics. I demonstrate how this combination preserves the insights of the original variably strict framework while combining it with the advantages of truthmaker semantics. This combination overcomes familiar problems concerning the interaction of conditionals and disjunctions such as Ross' paradox, free choice permission, and simplification of disjunctive antecedents. It furthermore provides a natural home for implementing the popular idea that modalized statements are grounded in modal anchors (entities in the world). This talk is based on joint work with Timothée Bernard (Université Paris Cité).

December 19th, 2024, Gail Weiss

Thinking Like Transformers - A Practical Session

With the help of the RASP programming language, we can better imagine how transformers---the powerful attention based sequence processing architecture---solve certain tasks. Some tasks, such as simply repeating or reversing an input sequence, have reasonably straightforward solutions, but many others are more difficult. To unlock a fuller intuition of what can and cannot be achieved with transformers, we must understand not just the RASP operations but also how to use them effectively.

In this session, I would like to discuss some useful tricks with you in more detail. How is the powerful selector_width operation yielded from the true RASP operations? How can a fixed-depth RASP program perform arbitrary length long-addition, despite the equally large number of potential carry operations such a computation entails? How might a transformer perform in-context reasoning? And are any of these solutions reasonable, i.e., realisable in practice? 

I will begin with a brief introduction of the base RASP operations to ground our discussion, and then walk us through several interesting task solutions. Following this, and armed with this deeper intuition of how transformers solve several tasks, we will conclude with a discussion of what this implies for how knowledge and computations must spread out in transformer layers and embeddings in practice.

December 12th, 2024, Yoad Winter

Understanding Partial Reciprocity

Reciprocal pronouns (‘each other’) and reciprocal verbs (‘meet,’ ‘hug’) allow non-maximal interpretations: both ‘the men fought’ and ‘the men fought each other’ can describe a barroom brawl where some men do not fight others. This talk connects non-maximal reciprocity to other cases of interpretative ‘slack’, especially plural definites (Križ & Spector 2020). Experimental findings show that while both intransitive and pronominal reciprocals allow non-maximality, intransitives are more tolerant towards exceptions, and are more influenced by agents’ intentions. Lexical theories of intransitives better explain these differences than quantifier-based approaches. Following Lasersohn’s (1999) view of slack quantification, we argue that pronominal reciprocals inherit ‘pragmatic halos’ from covert distributivity operators (Beck 2001), while intransitives derive their halos from slack relations with transitives, similarly to other lexicon-based semantic relations. Time permitting, I'll discuss the more general issues surrounding "strongest meaning" effects and homogeneity with reciprocals.

November 28th, 2024, Omri Amiraz 

Constraints on quantification: A partition-based approach

Quantifiers like most of are compatible with some predicates but not others, as in Most of the road is {narrow / \#long}. In the context of collective predicates, this contrast is often referred to as the gather/numerous distinction. Previous studies have argued that gather-type collective predicates allow distributive inferences to at least some non-atomic subparts of a plurality. However, this paper challenges that view, arguing that distributive-like entailments are not necessary for a predicate to be classified as gather-type. Focusing on the gather/numerous distinction in the context of subatomic quantification, it is proposed that a  gather-type predicate must permit a partition of the entity's subparts, induced by membership in the denotation of the predicate. Sentences violating this constraint lack determinate truth conditions, rendering them unacceptable.

November 21st, 2024, Noga Zaslavsky

Efficient data compression underlies meaning in language

Our world is extremely complex, and yet we are able to exchange our thoughts and beliefs about it using a relatively small number of words. What computational principles can explain this extraordinary ability? In this talk, I argue that in order to communicate and reason about meaning, both humans and machines must efficiently compress their representations of the world. In support of this claim, I present a series of studies showing that: (1) languages evolve under pressure to efficiently compress meanings into words; (2) the same principle can help reverse-engineer the visual representations that underlie human semantic systems; (3) efficient compression may also explain how meaning is constructed in real time, as interlocutors reason pragmatically about each other; and (4) these findings offer a new framework for studying how language may emerge in artificial agents without relying on human-generated training data. This body of research suggests that efficient compression underlies meaning in language and offers a cognitively-motivated approach to emergent communication in multi-agent systems.

November 7th, 2024, Michela Ippolito

What makes rhetorical questions rhetorical?

Most proposals about the nature of rhetorical questions involves the notion of common ground: for example Capoingro and Sprouse (2007) argue that a rhetorical question is a question whose answer is already in the common ground; Biezma and Rawlins (2017) argue that rhetorical questions are questions presupposing that their answer is in the common ground. In this talk, I will discuss two types of questions that I argue challenge what I call the common ground view of rhetorical questions: discordant questions (Ippolito 2021, 2024), and Exclamative-Interrogatives (Chen and Ippolito 2024). I will then draw a preliminary sketch  for a possible alternative view of rhetoricity. 

October 17th, 2024, Hartmut Fitz

Neuronal theory of memory for language processing

Language processing requires memory on different timescales, ranging from milliseconds to seconds and minutes. What is the nature of this "processing memory", and how is it implemented in the neurobiological infrastructure of the brain? In this talk I will first review structural theories of short-term memory that are prominent in neuroscience. Then I will describe an alternative approach, developed in my group, that locates memory in single-neuron dynamics. I will show that this mechanism provides memory that is context-dependent and sensitive to serial order which makes it suitable for sentence comprehension. Finally, I will highlight some open questions that theories of processing memory should address in the future. A background in neuroscience is not needed to follow my talk.

September 26th, 2024, Ido Benbaji

Missing temporariness implicatures and their consequence for the theory of alternatives

(Joint work with Zachary Feldcamp, MIT)

Abstract: We use the domain of English aspect marking to probe for the correct algorithm of alternative generation, providing a novel case of implicatures disappearing when syntax independently rules out the alternatives required to generate them. We argue that the data lends novel support to the structural theory of alternative (particularly in contrast to the focus theory).

More details: Our argument for the structural theory of alternatives rests on two empirical observations. First, an old observation by Dowty (1979) that certain present-tense stative predicates with immobile objects are infelicitous in the progressive aspect (1).

a. New Orleans lies at the mouth of the Mississippi.

b. #New Orleans is lying at the mouth of the Mississippi.

Dowty suggests that (1b) is infelicitous due to a temporariness inference, according to which New Orleans is only temporarily located at the mouth of the Mississippi, contra what we know about the permanence of cities' locations. The inference can be derived as an implicature by assuming that the imperfective example in (1a) is an alternative to (1b), and thus that by asserting (1b) one implies the negation of (1a).

The second observation, novel – as far as we know, is that the infelicity of the progressive disappears in particular syntactic environments that all happen to disallow the imperfective, like VP-fronting in (2).

a. Lying at the mouth of the Mississippi, is the city of New Orleans.

b. *Lies at the mouth of the Mississippi, is the city of New Orleans.

We show that the distribution of the infelicitous temporariness inference can be correctly captured if we assume that the alternatives involved in implicature computation are derived structurally, but not if they are derived by manipulating semantic objects alone.

September 5h, 2024, Uli Sauerland

On Movement Structures

A large class of linguistic structures are generally referred as involving "movement" and are frequently analyzed as involving a special relation of phrasal identity (copying or structure-sharing).  But I show that evidence from both adult and child language supports the view that such structures involve two independent phrases linked by a binding relationship. Therefore semantic compatibility, but not phrasal identity is required.  I then explore a view where binding relationships are primary, and movement dependencies are only formed in the pronunciation component of grammar.

June 13th, 2024, Pritty Patel-Grosz

A super linguistic analysis of electronic dance music

Recent years have witnessed the emergence of a body of work that extends tools broadly inspired by linguistic theory to other research domains, such as visual narrative or music. A central goal of this methodological extension is to attempt to identify cognitive mechanisms that are found to be at play across different modalities: communicative mechanisms that are shared by comics, music, gesture, and dance. In this talk, I argue that electronic dance music (EDM) (e.g. house, techno, garage and dubstep) exhibit a parallel structural organization to what has been proposed for comics without words after the model of hierarchical structure proposed in theoretical linguistics. This parallelism suggests that both systems are governed by general cognitive mechanisms for the narrative organization of tension and release, which are not modality-specific. By introducing linguistically informed analytical tools, further phenomena pertaining to the scientific analysis of EDM meaning can be investigated. I ague that the semantics of EDM is directive/imperative, and maps onto instructions to dance with varying levels of intensity. Finally, I discuss preliminary findings concerning the gestural modality, where we find that the communicative gestures among dancers appear to have a form and function that resembles the communicative gestures of non-human great apes.

June 6th, 2024, Christina Pawlowitsch 

Meaning in Costly-Signaling Games

Costly-signaling theory is a well-established paradigm in economics (Spence 1973) and theoretical biology, where it is mostly discussed under the name of the Handicap Principle (Zahavi 1975). It has been advanced as an explanation for a wide range of phenomena - from educational credentials as a signal for productivity in the job market (Spence 1973), over the evolution of 'handicaps' (Zahavi 1975, Grafen 1990), to the practice of inefficient foraging strategies as signals of social status (Bliege Bird et al. 2001, Bliege Bird and Smith 2005). Game theory provides a coherent modeling framework for costly-signaling phenomena by explicitly formalizing the interaction as a game with incomplete information. Signaling games have the notorious property that they have multiple equilibria. What makes costly-signaling games an interesting modeling tool from a semantic point of view is that the meaning of a signal is not given ex-ante as an assumption of the model, but arises endogenously, in equilibrium (or any other solution concepts that one wants to employ), as a function of the parameters of the model, notably, the cost of the signal for different types and the prior probability distribution over types. In this talk, I will investigate the properties of the equilibrium structure of costly signaling games from this semantic point of view. I will do so by looking at a parametrized family of games with two different types ('good' and 'not good'), two signals (the presence and the absence of a costly signal), and two different reactions to signals ('accept' and 'do not accept'). Equilibria will be discussed from a 'classical' game-theoretic point of view, using the concept of Bayesian Nash equilibrium, and from an 'evolutionary' point of view, focusing on the replicator dynamics. For each constellation of parameters, the entire equilibrium structure is captured. Instead of focusing on a particular kind of equilibrium (such as perfectly revealing equilibria), I will focus on how the equilibrium structure changes as a function of changes in the parameter specification of the model. The hypothesis I try to defend is that Meaning in these models is to be found not in one isolated equilibrium but in the entire equilibrium structure. The explanatory potential of this modeling framework is exemplified by two applications: style-shifting and politeness in language.

May 30th, 2024, Joshua Zami (animal communication)

Implicit expectations and iconicity of great apes' gestures in adult humans

Recent studies have shown that humans are capable of interpreting great ape gestures in line with the uses/meanings that have been proposed in the primatology literature at an above-chance level (Graham & Hobaiter 2023; Henderson et al. 2024). This gives rise to the question whether humans have implicit expectations regarding the meanings of ape gestures, and if iconicity is linked to these expectations. I discuss two ongoing experiments that aim to answer these questions. In Experiment 1, participants are asked to first give the meaning of different ape gestures performed by a human, and then rate the gestures’ iconicity on a 10-point scale. In Experiment 2, participants are presented with the same set of gestures accompanied by meaning suggestions; their task is to decide whether the suggested meaning is correct for the gesture it is paired with or not; participants’ response times and accuracy are recorded as the dependent variables; the aim is to see if participants are faster (and/or respond with more accuracy) in evaluating iconic gestures than non-iconic gestures.

May 23rd, 2024, Barbu Revencu

Not a Pipe: STAND-FOR Relations in Human Cognition

Local assignments from visual objects to discourse referents surface across many forms of human communication, such as pretend play, puppet shows, diagrams, and animations: A banana stands for a phone, a puppet stands for an agent. I call such assignments "STAND-FOR relations" and hypothesize, based on their ubiquity, that humans are endowed with a dedicated cognitive mechanism that allows them to interpret and represent these relations. I will present three studies that provide evidence for the operation of such a mechanism in both infants and adults. Study 1 offers an existence proof that 19-month-olds already interpret one class of stimuli—animations—as representations. Study 2 shows that 15-month-olds accept arbitrary visual shapes as symbols, which indicates that STAND-FOR relations are easily activated and available early in human development. Study 3 provides evidence from adult participants that even realistic photos of objects are interpreted as symbols. Taken together, the studies reveal medium–content dissociations in both infants and adults, across a variety of representational formats—a potential signature of STAND-FOR relations.

April 25th, 2024, Jennifer Culbertson (LINGUAE Lectures)

A developmental and cross-modal perspective on cognitive biases in language

Language has been argued to be shaped by two extremely general cognitive pressures: simplicity and expressivity. There is ample evidence to suggest that these two pressures can help explain why languages exhibit compositionality, systematic structure and regularity. These are key features that characterise how meaning is mapped to form in all domains of language---from the lexicon to syntax. However, the vast majority of this evidence comes from studies of spoken languages. In addition, almost all experiment work exploring these pressures has been conducted on adults. And yet, it is clear that the capacity for human language is multi-modal and changes across development. In this talk, I discuss recent and ongoing experimental work exploring the role of simplicity and expressivity in shaping communication systems across modalities and development.

April 24th, 2024, Jennifer Culbertson (LINGUAE Lectures)

Beyond typology: experimental explorations of language universals

Human languages exhibit striking variation. At the same time, certain linguistic patterns crop up again and again, while others seem to be extremely rare. What these tantalising observations tell us about human language is one of the most contentious questions in linguistics. Do similarities between languages reflect a special capacity for language that has evolved only in humans? Do they reflect more general features of the human mind, potentially shared with our ancestors? Are they just down to accidents of history? Traditionally, linguists have argued for one or another of these answers based on limited sources of evidence. For example, it is common to base claims on small samples of languages, case studies of how a handful of languages change over time, or examples of how individual languages are learned. In this talk, I highlight problems with these traditional sources of evidence and survey what I see as the solution: cross-cultural experiments. I show how this approach can be used to bring crucial empirical evidence to bear on how language is shaped (or not!) by the human linguistic and cognitive system.

April 18th, 2024, Yasu Sudo

Presuppositions of Strong (and Weak) Determiners

Why do some quantifiers presuppose their domain to be non-empty, and others not necessarily so? We propose an answer based on the interplay of dynamicity and a ban on idleness.

April 11th, 2024, Jad Wehbe

Covert reciprocals: A syntactic account of reciprocal alternations

Certain predicates likehug require an overt object when the subject is singular (1a), while being able to surface without an overt object when the subject denotes a plurality (1b). These alternations have been referred to as reciprocal alternations (Levin, 1993; Winter, 2019), and indeed it is tempting to paraphrase the 1-place variant in (1b) with a reciprocal object as in (1c). A straightforward account of this alternation would be to propose that (1b) has a covert reciprocal in object position (Hackl, 2002). A major challenge for this approach is that the 1-place variant and the overt reciprocal are not always truth-conditionally equivalent (Carlson, 1998 a.o.). For example, in the context in (2), (1c) is true while (1b) isn’t. The truth-conditional differences between the 1-place variants and overt reciprocals have led to the conclusion that the 1-place variants should instead be analyzed as 1-place collective predicates, where a predicate likehug is taken to be lexically ambiguous between a 1-place and a 2-place predicate (e.g. Winter, 2018).

  (1)   a. Jane hugged Mary.

          b. Jane and Mary hugged.

          c. Jane and Mary hugged each other.

 (2) Context: Jane hugged Mary while Mary was sleeping and then Jane fell asleep and Mary woke up and hugged her. (Winter, 2018)

The goal of this talk will be to (i) propose a covert reciprocal analysis that can account for the truth-conditional differences between overt and covert reciprocals and (ii) provide evidence that (1b) is in fact a reciprocal. I assume a decompositional analysis of reciprocals, following Heim et al. (1991) and propose that the differences in truth-conditions are due to a requirement that the reciprocal has to be bound as low as possible when it is elided. I argue that all of the differences between covert and overt reciprocals can then be accounted for as due to differences in the scope of reciprocal binding.  I show that this account correctly predicts that covert reciprocals, just like their overt counterparts, give rise to homogeneity, non-maximality, and cumulative readings, even when the subject denotes an individual of cardinality two, as in (1b). I argue that these parallels between the overt reciprocal in (1c) and the 1-place variant in (1b) provide strong evidence for a theory which treats (1b) as a covert reciprocal.

April 4th, 2024, Pierre Jacob

Reflections on mentalizing in human infancy

March 28th,  2024, Alexey Koshevoy

Convexity bias makes languages efficient

Peter Gärdenfors famously argued that the concepts are convex. Recent literature has presented compelling evidence that convex categories represent a learning advantage in humans, baboons, and neural networks. It was suggested that the ease of learning or language efficiency could explain the convexity of the human lexicon. However, it is unclear how communication benefits from the convex structure of the mental representation. In our talk, we will argue that convex categorization is the driving force that permits semantic categories to optimize the simplicity-informativeness trade-off. We will first present evidence that artificial color systems with the highest degree of convexity display the best optimization of these two pressures, which matches natural language data. Then, we will show evidence that communicating artificial agents optimize the simplicity-informativeness trade-off better when equipped with a convexity bias rather than agents biased towards simplicity or unbiased agents.

March 21st, 2024, Dean McHugh

Modality disordered

Orders play an essential role in current thinking about the semantics of modality. They lie at the heart of the similarity approach to conditionals (Lewis 1973) and more generally, Kratzer’s (1981) analysis of modals, which orders possibilities along various dimensions (such as similarity, desirability, and plausibility). It is a remarkably general, compelling picture. However, in this talk we present evidence that it is mistaken for would. That is, the semantics of would cannot be expressed in terms of an order over possibilities—no matter how the order is intuitively understood.

The arguments all involve Reciprocity, the inference from A > C, A > B and B > A to B > C, where A > C denotes “if A, would C”. On the similarity approach to conditionals and Kratzer’s (1981) semantics of would, Reciprocity follows from reflexivity and transitivity of the order—the bare minimum constraints to impose. We present three pieces of evidence that Reciprocity is invalid. Firstly, experimental evidence in which half the participants judged the rule invalid (joint work with Tomasz Klochowicz). Secondly, new counterexamples to a consequence of Reciprocity, called Substitution: that whenever A and B are logically equivalent, then so are A > C and B > C. (Failures of Substitution are especially interesting since they show the limits of a purely truth-conditional approach to meaning.) Thirdly, that Reciprocity makes the wrong predictions for the semantics of causal claims, assuming a counterfactual approach to causation.

If not with orders, how should we model the meaning of would? This talk proposes a new approach based on aboutness. The idea is that when we interpret a would-conditional, we (i) pick a time at which to imagine a change, (ii) vary the part of the world its antecedent is about at that time, and (iii) play the laws forward. We show that this approach overcomes the problems of ordering approaches.

March 7th, 2024, Catherine Crockford

Primates: the emergence of compounding

Only one natural communication system is considered combinatorially complex with respect tomapping complex structure to complex meaning. This is human language. With alimited sound set, we use a large variety of compounding mechanisms toendlessly generate both new and relational meanings. Most animals have alimited sound set, and per species may only have one or no compoundingmechanism at their disposal. However, recent studies delving deeper into vocalsequence production suggest animal vocal compounding capacities have beensubstantially underestimated, particularly in one of our closest livingrelatives, the chimpanzee. Chimpanzees may have a limited capacity to inventnew calls but demonstrate highly flexible abilities to combine calls, such thatmost call types combine with most others. Vocal sequences show ordering and recombinatorial properties, withpopulation differences suggestive of social learning capacities. Like humans,but unlike African monkeys, ontogenetic development is protracted, withutterance length and diversity dramatically increasing until 10 years of age.Vocal sequences fit patterns predictive of encoded meaning. Whilst there hasrightly been a focus in animal communication studies on signal comprehension,given the unfolding observed vocal sequence complexity, it is timely to embracequantitative assessment of the potential of vocal production.

February 29th, 2024, Matthew Mandelkern

Diamonds are a disjunction’s best friend

I introduce a number of new puzzles about the relation between disjunction and possibility. I argue that these puzzles are best solved with a theory on which 'p or q' means (p∨q)∧<>p∧<>q, where <> is a possibility modal underspecified for modal flavor. I show that the resulting theory also yields an elegant account of free choice inferences.

February 15th, 2024, David Omer (animal communication)

The Naming of Non-Human Primates 

Humans and dolphins are the only known species which vocally label other conspecifics.Until now, it was unknown if non-human primates possessed a similar language-basedability. We recorded spontaneous phee call dialogues between pairs of marmoset monkeys and discovered that marmosets use phee calls to convey their location and to vocally labelother conspecifics. Additionally, they respond more reliably to calls intended for them.Analysis of calls from various monkeys highlighted family group structures, with membersof the same family using similar acoustic feature patterns to address others and performvocal learning. This discovery illuminates the complexity of social vocalizations in non-human primates and suggests that marmoset vocal communication could serve as a modelfor human language, offering insights into social evolution that were previously unknown.

February 8th, 2024, Alex Kalomoiros

Hurford phenomena: a theoretical and experimental investigation

Fifty years ago, Hurford 1974 proposed the constraint whereby a disjunction is odd if one disjunct entails the other:

(1) # John visited Paris or he visited France.                  

Ever since, these types of disjunctions have proven instrumental in the theorization of various phenomena, from redundancy (c.f. Schlenker 2009, Katzir & Singh 2013) to implicature (c.f.Chierchia, Fox & Spector 2012). However, a unified account of Hurford phenomena is still lacking (cf. Marty & Romoli 2022). The present talk aims to make progress towards such a unification.                 

The theoretical component of the presentation aims to provide a unified account of Hurford Disjunctions like (1) above, and Hurford conditionals (Mandelkern & Romoli 2018), like (2)below:                  

(2) # If John is not in Paris, he’s in France.                  

The puzzling thing about such conditionals, is that while (1) and (2) are equivalent under theor-to-if tautology of classical logic (and hence Hurford’s constraint might be expected to apply to both), (3) and (4) are also equivalent under the same tautology. But while (4) sounds felicitous, (3) is predicted to be odd under Hurford’s constraint:                  

(3) (#) Either John didn’t visit France or he didn’t visit Paris.(4) If John visited France, he didn’t visit Paris.                 

We pursue a unified account by arguing that in fact we can find contexts where (3) sounds natural and proposing a new notion of redundancy (dubbed super-redundancy) that penalizes Hurford disjunctions and conditionals like (1) and (2), but not sentences like (3) and(4).                 

The second part of this talk aims to directly test in an experiment the predictions made by an account based on super-redundancy. While the results offer support for some of the predictions of the super-redundancy approach, they also open up new puzzles, and as such the talk concludes with some suggestions on future directions aiming to tackle these puzzles.

January 18th, 2024, Cornelia Ebert & Marcus Steinbach

Lexical Demonstrations in Spoken and sign Languages

Typological research has shown that many typologically different spoken languages haveideophones, a special class of words which can be defined as “an open lexicalclass of marked words that depict sensory imagery” (Dingemanse 2019). Based ondata from German Sign Language (DGS), we argue that also sign languages have aspecial class of signs, so-called idiomatic signs, that share the definingproperties of ideophones and show related semantic behavior. Building onco-speech gesture semantic approaches (Ebert et al. 2020) and applications thereofto ideophones (Barnes & Ebert 2023), we suggest an account thatcompositionally derives the meaning of these idiomatic signs. We show that ideophonesas well as idiomatic signs consist of a conventionalized at-issue meaning partand a gestural non-at-issue part, whose meaning can vary from utterancesituation to utterance situation and is dependent on how the ideophone/idiomaticsign is produced in the given situation. We hence argue that sign languagesjust like many spoken languages have a special openclass of marked lexical expressions that depict sensory imagery. Unlikeideophones in spoken languages, however, the depiction of idiomatic signs isnot auditory but visual. And unlike other iconic lexical expressions in signlanguages, idiomatic signs are depictive context-dependent expressions with aparticularly expressive meaning component, that is, the iconic features(especially the iconic nonmanual ones) typically contribute an expressivedepictive meaning component that is dependent on the very performance of thesign in a specific situation. 

January 11th, 2024, Timothée Bernard

Studying the syntax and semantics of emergent languages

In signaling games, two or more agents (human or otherwise) have to cooperate through information exchange in order to retrieve some target. In particular, we start with an experimental setup in which one agent has to describe an image using arbitrary symbols so that another agent can recognise it among a set of distractors. The agents are implemented as neural networks and their training is based on reinforcement. Our long term goal is to understand under which conditions the language developed by the two agents, similarly to natural language, exhibits a syntax and a compositional semantics. As first steps, (i) we study how many implementation choices affect the reliability of the training procedure and the performance of the agents, (ii) we introduce an array of automatic metrics designed to uncover syntactic and semantics aspects of the messages exchanged by successful pairs of agents and (iii) we show that these pairs of agents can acquire and communicate about high-level concepts even when apparently simpler strategies are available (and indeed discovered).

December 21st, 2023, Carlo Geraci & Sharon Peperkamp

Word recognition in LSF*: The relative importance of different phonological parameters

November 23rd, 2023, Filipe Kobayashi 

From NPs to predicates of individual concepts

The main goal of this project is giving a principled account of a puzzle identified by Nathan (2006).  On the one hand, sentence (1) can be true even if the set of governors remains constant throughout the relevant interval of evaluation -- it is true as long as no state has the same governor as before.  On the other hand, sentence (2), whose structure is superficially identical to (1), can only be true if every book on my shelf has been replaced by a book that was not on my shelf before.  Nathan observed that the relevant distinction between (1) and (2) is that the NP in the first sentence is headed by a relational noun.

(1) Every governor changed.

(2) Every book on my shelf changed.

I propose a novel solution to this puzzle by positing a unified procedure for shifting NPs from properties of individuals into predicates of ICs.  This procedure bears strong resemblance to Kartunnen's (1977) semantic analysis of interrogative clauses: NPs are turned into suitable predicates of ICs through the interaction of a type-shifting operation and existential quantifiers.  Under this proposal, the difference between (1) and (2) will follow from the different internal structures of the NPs involved.

October 19th, 2023, Johanna Mesch

Methodological issues in analyzing tactile signed interaction

This presentation reflects on the challenges involved in creating data collection and analyzing interactions in tactile sign language. It is based on personal experiences in the field of research over time. The presentation begins by discussing conversational data between tactile signers with acquired deafblindness, using Conversation Analysis methods to describe the interaction of deafblind signers, particularly focusing on turn-taking mechanisms and intersubjectivity.

October 5th, 2023, Jean-Remy Hochmann

Incomplete Language of Thought in infancy

The view that infants possess a full-fledged propositional Language of Thought (LoT) is appealing, providing a unifying account for infants’ precocious reasoning skills in several domains. However, careful appraisal of empirical evidence suggests that infants may lack a crucial component of a propositional LoT: discrete representations of abstract relations.

September 28th, 2023, Pritty Patel-Grosz (Animal linguistics)

Two puzzles in animal linguistics: formal semantics of ape gestures and the syntactic properties of dance displays in crested gibbons

A growing body of research applies methodology inspired by formal linguistics to body-based meaning making in our closest primate relatives. In this talk, we address two types of communicative body movement in primates, great ape gestures and gibbons dance.

First, we probe the question of whether human and non-human great apes (humans, chimpanzees, bonobos, gorillas and orangutans) share a common set of directive (interactive) gestures and their meanings. In an exploratory experiment (n=300), we tested whether human participants would infer the same gesture functions that have been observed in chimpanzees. Notably, a large number of ape gestures are multifunctional, in that they are observed to communicate one function in some contexts and a different function in other contexts. Applying the tools of formal semantics, we find that an abstract core meaning can be proposed for such multifunctional gestures based on the overlap between the attested functions (e.g. between a “Stop that” function and a “Move away” function); our search for a hypothesized inventory of universal primate gestures needs to look for meanings at this level of abstraction, rather than a more concrete level.

Second, we present findings from close-range video recordings of adult females of four crested gibbons species in captivity (N. annamensis, N. gabriellae, N. leucogenys, & N. siki). Our results demonstrate that dances in Nomascus represent a common and intentional form of visual communication. A linguistically informed view of this complex sequential behaviour demonstrates that gibbon dances follow a nested grouping organization, a pattern so far not described for visual displays in other non-human primates.

September 21st, 2023, Will Merrill

Entailment Semantics Can Be Extracted from an Ideal Language Model

Language models are often trained on text alone, without additional grounding. There is debate as to how much of natural language semantics can be inferred from such a procedure. We derive an entailment test that provably detects entailment between pairs of sentences based on their co-occurrence probabilities assuming the sentences are generated by Gricean agents, i.e., agents who follow fundamental principles of rational communication. We show how this test can decode entailment judgments from language models trained on synthetic Gricean data. Finally, we use the test can also be used to decode entailment judgments from language models trained on natural language and explore the implications of these empirical results for theories of pragmatics. Our results suggest pragmatics is a link that enables semantic information to be extracted from raw text (at least to some extent) and that empirical data from language models may be useful for testing and developing computational pragmatics theories.

July 6th, 2023, Megha Sundara

How infants discover morphological suffixes and use them to discover phonemes: Experimental and Computational findings

Morphological suffixes like -s, -ed and -ing are the basic building blocks of English syntax. In this talk I will present research to identify the representations, time course and the mechanisms involved in the acquisition of morphology. To do so, I will integrate looking time experiments with infants (n=~350), corpus analysis and computational modeling.

    I will show that English-learning infants begin to discover suffixes before 6-months. And they do so without access to meaning, or function. Additionally, I will show that infants use morphological suffixes to discover phonemes. The early discovery of suffixes is problematic for whole word models of acquisition where morphology is a by-product of shared form and meaning (Baayen et al., 2015; Tomasello, 2000). Instead, I will argue that infants represent morphemes from the earliest stages of acquisition, and that morphological acquisition precedes the discovery of phoneme (and word class).

    Next, I will use computational modeling to provide a formally explicit account for how these suffixes could be learned. Model performance will be evaluated against the developmental trajectory of morphological acquisition uncovered in the infant experiments. Then, I will use the computational model as a guide to pinpoint infant experiments likely to distinguish alternate theories of morpheme discovery. Finally, I will set up predictions for the trajectory of morphological acquisition in infancy in French, Spanish and German.

June 29th, 2023, Catia Correia-Caeiro (Animal linguistic communication)

Perception and production of visual cuesacross species

Visualcues such as facial expressions or body gestures are central for communicationand emotion expression in social species. Traditionally, facial expressions havebeen studied in a gestalt way or using subjective labels of emotion. Thisintroduces bias in the research questions and misses subtle but important cues,particularly when looking at visual cues of non-human animals. Hence, part ofmy work has been to develop objective tools to precisely measure facialexpressions in a range of species, including primates and domestic animals.These tools, the AnimalFACS (AnimalFACS.com), are adapted from the humanFACS (Facial Action Coding System), which avoids issues of a priori assumptionsof meaning, and allows comparisons between species solely based on the anatomyof the face. FACS tools are used to investigate the function and evolution ofcommunication and emotion expression in human and non-human animals, providingnovel insights into each species social behaviour. I will briefly explain howFACS works and then present an overview of my studies that have applied FACS toinvestigate both perception and production of facial expressions in combinationwith other cues modalities (i.e., body gestures and vocalisations). Togetherwith FACS, I use non-invasive and cooperative methods (e.g., eye-tracking, salivaryhormones, cognitive paradigms) to investigate a range of topics, such asintentionality of facial expressions, perception of full body emotion cues, andhormonal responses to visual cues. By using objective tools to investigatevisual cues in a comparative and evolutionary way, we can not only betterunderstand human and other animals' evolution of communicative and emotionalprocesses, but also directly apply this knowledge to improve species welfareand human-animal relationships.

June 22nd, 2023, Thierry Lengagne (Animal linguistic communication)

Systèmes de codage et décodage des informations par les animaux dans différents contextes comportementaux 

Thierry Lengagne est chercheur au CNRS en écologie comportementale. Il est spécialisé sur la problématique de la communication animale et a travaillé sur plusieurs dizaines d’espèces animales en milieu naturel (oiseau, mammifère, amphibien) dans différents contextes comportementaux comme celui de la reconnaissance individuelle de la sélection sexuelle ou de la communication anti-prédateur. Il a souvent cherché à prendre en compte l’ensemble de la chaine de transmission de l’information émetteur-canal-récepteur afin de bien comprendre les différentes contraintes évolutives pesant sur les systèmes de communication. La présentation de 22 Juin sera centrée sur la partie de son travail concernant les systèmes de décodage de l’information par une approche expérimentale qui a fait l’objet de plusieurs dizaines de publications.

June 15th, 2023, Brent Strickland

"Myside" biases and motivations at organizational scale

The "Myside" bias, i.e. a disposition to systematically search for and believe information that corresponds to one’s longstanding beliefs, is one of the most robust cognitive biases known to cognitive psychologists. It is operational across cultures and age groups. Typically this bias and associated “myside motivations” have been studied at an individual level. In the current presentation I examine the costs and possible benefits of myside bias at an organizational/societal scale. In particular, recent work from my group has illustrated some surprising costs of myside bias. These include subtle reasoning errors around societally divisive topics, a propensity to treat people who fail to show “appropriate bias” as disloyal, and a disposition to overshare “mysided” information that touches on core moral values. On the other hand, myside bias is also associated with some surprising benefits. Particularly while groups with a higher degree of myside bias are more error prone (clearly bad), they are also faster to make decisions. A better understanding of the tradeoffs associated with bias can help to better anticipate and evaluate the consequences of changes to incentive structures or procedures shaping collective practices.

June 1st, 2023, Roni Katzir

On the roles of anaphoricity and questions in free focus

The sensitivity of focus to context has often been analyzed in terms of anaphoric relations between sentences and surrounding discourse. Data from Wagner (2006) and Katzir (2013) challenge earlier anaphoric accounts of focus placement and raise the possibility of a non-anaphoric account in which questions play a central role, a direction pursued by Büring (2019). However, while an account without special felicity conditions on focus placement seems preferable on conceptual grounds, there are cases where the fine control offered by anaphoricity seems needed, either to block deaccenting that would be licensed by a question or to allow local deaccenting that is not warranted by a question. Such cases challenge non-anaphoric accounts and appear to support recent anaphoric proposals such as Schwarzschild (2020) and Goodhue (2022). I argue that these potential arguments for anaphoricity are only apparent and that to the extent that the anaphoric component is not inert it is in fact harmful. I propose that the distribution of focus is not governed by any direct felicity constraints, anaphoric or otherwise. Rather, what determines whether focus placement is felicitous is indirect: focus feeds semantic processes such as exhaustification and question formation, and the result should be an acceptable contribution to discourse.

May 31st, 2023, Roni Katzir

Questions, answers, and the typology of logical connectives (joint work with Moysh Bar-Lev)

Why are AND and OR the only binary connectives that are lexicalized as simplex? Horn (1972), who first observed this striking typological fact, suggested an account that relied on communicative considerations, and in particular on the strengthening of utterances in conversation through scalar implicature. Much subsequent work on Horn’s pattern has adopted this idea, and we will make use of it in this talk. We start from the observation in Bar-Lev & Katzir (2022) that AND and OR are the only connectives for which exhaustification excludes alternatives that are logically independent of the connectives’ arguments. We consider this observation in light of a particular view on communicative success according to which it is better if a speaker's utterance can convey their epistemic attitude towards each cell in the question under discussion. We show that on this view, AND and OR are the only connectives that result in successful communication when logically-independent alternatives are under consideration. Moreover, we show that this communicative advantage of AND and OR would be lost if they had other connectives as alternatives; this paves the way for explaining not only why lexicalization of AND and OR is very common cross-linguistically, but also why other connectives are never lexicalized as simplex.

May 25th, 2023, Janek Guerrini 

Revisiting kind predication

In this work, I present novel data form Italian, showing that genericity interacts with mood. Definite plural generics may receive a law-like or an accidental flavor when the subject is modified by a relative in the indicative. However, when it is modified by a relative in the subjunctive, it can only receive a law-like reading. I propose that this is due to a structural ambiguity in Italian definite plural generics. Assuming that definite plurals can denote kinds, and that kinds are plural entities, we expect the distributive operator DIST to operate on kinds just like it does on sums denoted by referential plurals. Its optional insertion then gives rise to two structures. (i) If DIST is not inserted, the kind is interpreted in the restriction of GEN, and we get the usual LF. The modal nature of this structure yields the law-like reading, and licenses the subjunctive. (ii) If it is inserted, it distributes the predicate over actual members of the kind, yielding the accidental reading. The subjunctive is then not licensed, as it cannot be interpreted in the modal environment provided by the restriction of GEN. This also predicts that singular indefinite generics cannot receive accidental readings, as they don’t denote kinds. I finally argue that a similar reasoning may provide a fresh perspective on English.

May, 11th, 2023, Amir Anvari

Origins of Conservativity in the Breakdown of the Projecting Mind

While it is well-established, apparent counter-examples notwithstanding, that natural language determiners denote conservative functions(Keenan & Stavi 1986, a.o.), an explanation of this Conservativity Universal (CU) has proven elusive despite intensive research. I will first offer some reflections to the effect that CU cannot be explained on grounds of simplicity or learnability (in line with recent findings, e.g. Steinert-Threlkeld and Szymanik2019, van de Pol et al 2023). Isolating a class of conservative/non-conservative minimal pairs, I will argue that functional approaches cannot break the tie between these pairs without stipulating an asymmetry. I will therefore consider an approach to CU based on the logical syntax of quantification, namely copy theory of movement. Extant work in this direction (Romoli 2015) has surprisingly argued that, contra CU, non-conservative determiners may very exist but we cannot be sure as, within copy theory, such determiners are bound to yield the same truth-conditions as conservative ones. This result stems from a tacit abstraction from principles of presupposition projection. I will argue that, with an auxiliary assumption about the denotation of determiners, an implementation of Schlenker’s (2009) projection theory coupled with Fox’s (2002, 2003) rule of trace conversion correctly predicts all determiners to be conservative. Time permitting, I will stress-test the proposal by considering local accommodation, the status of CU beyond determiners, and Strong Kleene Projection. I will finish by observing that, under certain assumptions, the same results can be secured even without copy theory, putting all the explanatory pressure on principles of projection. 

April 13th, 2023, Stefano Palminteri

Studying and improving reasoning in humans andmachines 

In the present study,we investigate reasoning in a large language model (GPT-3) with tools derivedfrom cognitive psychology traditionally used to demonstrate bounded or biasedreasoning.  Our results show that GPT-3presents cognitive limitations similar to those elicited in humans using thesame tasks, which are popularly captured under the “system 1” umbrella term. Strikingly we also show that these biases increase with the complexity of the model (andare not explained by the incapacity of performing mathematical tasks of equal computational complexity).  We also show that ispossible to devise strategies to "prompt" better performance in bothsystems, but the same strategies do not necessarily work in humans and machines. Finally,we conclude by discussing the epistemological and ethical implications of the rising field of AI psychology.

April 6th, 2023, Nina Haslinger

Unifying homogeneity and implicatures: The road less traveled

Definite plurals in upward-entailing environments tend to pick out the maximal plurality in the domain provided by the NP (1-a); in downward-entailing environments, their default interpretation is a weaker one that quantifies existentially over the NP-domain (1-b). These so-called homogeneity effects (see e.g. Schwarzschild 1993, Löbner 2000, Križ 2015) are often accounted for in terms of lexically triggered semantic gaps between the truth and falsity conditions.

(1)

a. The switches are on.

default interpretation: `All the switches are on.’

b. It’s not the case that the switches are on.

default interpretation: `None of the switches are on.’

An alternative view pursued in the recent literature (e.g. Magri 2014, Bar-Lev 2018, 2021, Chatain 2021, 2022, Hirsch & Schwarz 2022) attributes the maximal interpretation of (1-a) to an implicature. If so, definite plurals are sensitive to monotonicity properties for the same reason that *some* gets strengthened to ‚some but not all‘ in (2-a) but not (2-b): In both cases, the strengthening is due to the presence of exhaustifying (Exh) operators that are either absent or have no effect in downward-entailing environments.

(2)

a. Anne read some of the assigned books.

default interpretation: `Anne read some, but not all of the assigned books.’

b. It’s not the case that Anne read some of the assigned books.

default interpretation: `Anne didn’t read any of the assigned books.’

These implicature accounts of homogeneity effects rely on two core intuitions:

- (i) that there are far-reaching parallels between maximal construals of plurals and strengthened construals of standard implicature triggers, so that a uniform approach is called for;

- and (ii) that this uniform approach should be implemented by analyzing both phenomena in terms of a uniform Exh operator.

In this talk, I develop an alternative view that accepts (i), but rejects (ii): Instead of deriving homogeneity effects with the theoretical tools developed for implicatures, I reanalyze standard implicatures in terms of the tools developed within implicature-less approaches to homogeneity.

I will start by summarizing some of the linguistic evidence in favor of (i) and then discuss two empirical problems for (ii), based on a) homogeneity-like effects in scalar predication (Feinmann 2022; Haslinger & Paillé, to appear) and b) contextual manipulations that make the homogeneity gap disappear (Augurzky et al., to appear). I then introduce a variant of the supervaluationist approach to homogeneity (Križ & Spector 2021) and show how to extend it to standard implicatures like (2). In the resulting system, implicatures give rise to truth-value gaps, as argued independently by Bassi et al. 2021, Paillé 2022 a.o. In addition to accounting for problems a) and b), the system captures the interpretation of definite plurals in the scope of an implicature trigger, as well as the *some*-over-*some* puzzle (see e.g. Gotzner & Benz 2022). I conclude by discussing an open problem for my approach, namely how to account for contrasts between non-monotonic and downward-entailing embedding contexts.

March 9th, 2023, Benjamin Spector

The confirmation paradox, Lewisian Imaging and the de re reading of restrictors (based on joint work with Vincent Mouly)

This talk is about the psychological side of the confirmation paradox.

Given a statement S of the form 'All As are Bs', people are more prone to think that an observation of an object that has both properties A and B 'confirms' S than they are to think that observing an object that is both not-B and not-A confirms S, even in a context where, from a normative point of view, these two types of observations should have the same confirmatory value (where confirmation is construed in Bayesian terms: o confirms S if P(o|S) > P(o)).

I will discuss a theory that explains this bias, which is based on the following assumptions:

- Restrictors of determiners are preferably interpreted de re

- Belief update proceeds by Lewisian Imaging rather than by Conditionalization (at least when assessing confirmatory value, cf. next point)

- o is perceived to confirm S given a certain belief state modeled by a probability distribution P if the probability distribution P_S that results from revising P with S by Lewisian Imaging is such that P_S(o) > P(o)

I will present preliminary results from an experimental study which support this theory, as well as additional thought-experiments (essentially experiments yet to be run) for which my introspective judgments also go in favor of the proposed theory. I will also discuss potential objections to the theory.

March 2nd, 2023, Olivier Morin

The puzzle of ideography

An ideography is a general-purpose code made of pictures that do not encode language, which can be used autonomously – not just as a mnemonic prop – to encode information on a broad range of topics. Why are viable ideographies so hard to find? I contend that self-sufficient graphic codes need to be narrowly specialized. Writing systems are only an apparent exception: at their core, they are notations of a spoken language. Even if they also encode non-linguistic information, they are useless to someone who lacks linguistic competence in the encoded language or a related one. The versatility of writing is thus vicarious: writing borrows it from spoken language. Why is it so difficult to build a fully generalist graphic code? The most widespread answer points to a learnability problem. We possess specialized cognitive resources for learning spoken language, but lack them for graphic codes. I argue in favor of a different account: what is difficult about graphic codes is not so much learning or teaching them as getting every user to learn and teach the same code. This standardization problem does not affect spoken or signed languages as much. Those are based on cheap and transient signals, allowing for easy online repairing of miscommunication, and require face-to-face interactions where the advantages of common ground are maximized. Graphic codes lack these advantages, which makes them smaller in size and more specialized.

This presentation will be based on the following paper, and the comments that it has received: https://hal.science/ijn_03837072

February 23rd, 2023, Isabelle Charnavel

Appealing to superlative clauses(or how to split the scope of superlative adjectives across intensional verbs)

The general goal of the talk is to argue for the hypothesis that just like comparative morphemes (e.g. English -er), superlative morphemes (e.g. English -est) can take a degree clause as argument, which denotes the domain of comparison. The specific goal consists in showing that this hypothesis provides a novel solution to two controversial empirical puzzles: how to derive (i) the low reading of intensional superlatives (e.g. the reading of the longest book that John said that Tolstoy had written when John – not the speaker – evaluates the lengths of the books under consideration), and (ii) upstairs de dicto readings (e.g. the reading of John wants to climb the highest mountain when John has a specific desire of climbing achievement without any comparative content). In both cases, the superlative clause hypothesis solves previously unsolved problems because it allows split scope between -est and the gradable adjective across the intensional predicate, as well as (variously partial) ellipsis of the degree clause.

February 16th, 2023, Lisa Hofmann

Anaphoric Accessibility with Flat Dynamic Update

A proceedings paper about the topic.

February 9th, 2023, Valentine Hacquard

How babies acquire clause types and their functions

If we hear a speaker ask “Where is Alex?”, we typically assume that the speaker doesn’t know where Alex is, and is seeking that information from their addressee. If we overhear a speaker say “Alex is home”, we typically assume that that speaker knows where Alex is, and is sharing that information with their addressee. What seems to drive our assumptions is the fact that, as competent speakers, we’ve come to associate interrogative clauses with questions, and declarative clauses with assertions. This talk investigates when and how children figure out this association: when and how do children learn the syntactic makeup of the clause types of their language and their canonical function, that is, the speech act that these clauses are typically used for?

We hypothesize that the mapping between clause type and speech act is a crucial precursor to the acquisition of syntactic features that emerge before age two, and that infants figure out form and function in tandem, with their growing knowledge of each mutually constraining the other. I will report on some experiments suggesting that children have linked interrogatives with questions and declaratives with assertions by 18 months, and on corpus analyses and computational modeling of child-directed speech showing how learners can draw from pragmatic and syntactic information and the way they interact, to figure out this mapping.

Februray 2nd, 2023, Killian Martin and Valerie Dufour (Animal linguistics communication)

Vocal complexity of rooks (Corvus frugilegus)

Why do some birds produce complex songs? The social complexity hypothesis for communicative complexity predicts a positive correlation between complexity in social organisation and complexity in the communication system of species, and may provide an explanation. Corvids are among the birds best known for their social complexity and high cognitive abilities. However, their vocal abilities remain generally poorly known, apart from their mimicking abilities. I developed tools for assessing vocal complexity, from a system for extracting and attributing vocalisations to the individual emitter, to a protocol for measuring vocal complexity directly from the spectrograms. These tools were developed and tested by studying the vocal complexity of rooks. This species exhibits a high degree of vocal complexity, based in particular on pronounced idiosyncrasies in the vocalizations of males. I also uncovered a capacity for spontaneous vocal entrainment to a rhythmic stimulus in this species, indicative of a control over the rate of song production. Moreover, the song produced by this species appears highly variable both in terms of vocal units and arrangement thereof, complicating formal analysis. These results support the social complexity hypothesis for communicative complexity, but also raise the question of the communication code between different individuals in this species. Highly individualised signals may require complex vocal learning to be understood by other individuals.

January 26th, 2023, Jan Sprenger

Counterfactuals: Semantics, Logic and Probability

The literature on the semantics of counterfactuals is divided between analyses as strict conditionals (e.g., C.I. Lewis) and variably strict conditionals (e.g., D. Lewis). In our paper, we first develop an alternative semantics for Lewis's counterfactual connective and characterize it using the Stalnaker conditional and the strong modal operator. This result reconciles the two predominant research programs on counterfactuals to a certain degree. Second, we axiomatize Lewis's counterfactual logics C0/C1/C2 in terms of the axiom system for the Stalnaker conditional. Third, our semantic analysis yields that the probability of a counterfactual corresponds to the belief function (in the technical sense of Dempster and Shafer) of the consequent, imaged on the antecedent. This is a notable generalization of Lewis's (1976) classical result for the Stalnaker conditional, and it has repercussions on how we should perceive the meaning of counterfactuals on the semantic levels. It is also a substantial achievement in the research program for an integrated semantics and epistemology of counterfactuals, whose probabilistic dimension has been studied much less than the probability of indicatives.

January 19th, 2023, Omri Doron and Jad Wehbe

A constraint on presupposition accommodation

Given the availability of presupposition accommodation, there is a challenge in identifying whether there is a difference between accommodating a presupposition and updating the common ground by accepting an assertion. Assuming a satisfaction view of presuppositions, where they impose conditions on the common ground and where accommodation involves moving to a new common ground, conditions on assertions such as anti-triviality have to be evaluated with respect to the updated common ground after accommodation. In the first part of the talk, we show, building on work by Irene Heim and Orin Percus, that this allows us to detect an asymmetry between accommodation and assertion in the case of global accommodation. We present novel data supporting this prediction. In the second part of the talk, we argue that this asymmetry is maintained with local accommodation, where locally accommodating a presupposition is unlicensed in certain cases if the asserted component does not contribute new information. We discuss the implications of this for different theories of local accommodation.​ 

December 15th, 2022, Jeremy Kuhn and Mora Maldonado 

Generalizing Polarity Sensitive Items to Positive Environments in an Artificial Language

Two well known diachronic patterns involve polarity sensitive items. First is Jespersen’s cycle: a polarity sensitive item appearing in the scope of negation is reanalyzed as itself the marker of negation. Second, many NPIs in English also have a secondary positive use, with a systematic semantic relation to the polarity sensitive use: in each case, the meaning of the positive form is the logical dual of the NPI. The present study investigates these meaning changes using an artificial language learning paradigm, where learners are trained on ambiguous input, and must generalize beyond this input. We ask which generalization strategies appear when learners have to interpret polarity sensitive items in a positive environment.

December 7th, 2022, Cedric Girard-Buttoz (Animal linguistics communication)

Complexity and structure of vocal communication in wild chimpanzees

 Human language exceptional combinatorial properties enabling endless meaning generation is in stark contrast to the very limited vocal sequence production capacities of non-human primates. This contrast led some researchers to question the suitability of vocal communication study on non-human primates to understand the evolution of language. However, studies have rarely quantified flexibility and structure of vocal sequence production at the whole repertoire level. Here, we characterised the structure of vocal sequences across the vocal repertoire of wild chimpanzees in the Taï population. We found that chimpanzees produced 390 unique vocal sequences. Most vocal units emitted singly were also emitted in two-unit sequences (bigrams), which in turn were embedded into three-unit sequences (trigrams). Bigrams showed positional and transitional regularities within trigrams, such that certain bigrams predictably occurred either in head or tail positions in trigrams, and predictably co-occurred with specific other units. We then assessed the potential for vocal usage learning of these sequences by comparing the call ordering of two single call the long-distance contact call pant-hoot and the submissive call pant-grunt into the greeting hoot sequence between two chimpanzee populations (Tai and Budongo). We found that both populations consistently differed in call ordering with pant-hoots being emitted first in the sequence in Taï and pant-grunt first in Budongo. These differences may be linked to different social pressures related to intra-group aggression and killing. Finally, we characterised the developmental trajectory of vocal sequence acquisition throughout ontogeny. This trajectory followed key developmental milestones with the steepest increase in the diversity and length of vocal sequences around the weaning age and the adult level of vocal complexity reached at maturity. Our results demonstrate a flexible complex vocal system in wild chimpanzees, potentially socially learnt. The capacity to organize single units into structured sequences offers a versatile system potentially suitable for expansive meaning generation. Further research must show to what extent these structural sequences signal predictable meanings.

December 1st, 2022, Márta Abrusán

Projection and the meaning of negation

A large array of linguistic content seems to 'project', i.e. escape the scope of an entailment-canceling operator such as negation. Why? I propose that the non-classical understanding of negation as a restricted modal operator (cf. Berto 2015) allows us to explain various projection effects. As a result, there is no need to assume that there is a natural class of 'projective content', any semantic content that ends up in the restriction is projective. In the second half of the talk I discuss case studies as well as the possibility of extending this idea to other entailment-canceling contexts.

November 24th, 2022, Damian Szmuc

Exploring all the three-valued presentations of classical logic

Cobreros, Egré, Ripley, and van Rooij have shown that the logic defined using the strict-tolerant consequence relation over the (three-valued) Strong Kleene schema coincides with classical logic (at least inferentially speaking). More recently, it has been proved by Szmuc and Ferguson that the (three-valued) Weak Kleene schema serves that purpose also. The aim of this talk is to generalize these results and present all of the three-valued schemata such that the strict-tolerant consequence relation over them coincides with classical logic. Finally, it will also be shown that these results have a connection with the proof-theoretical search for all the calculi appropriate for classical logic, where certain formal requirements are met. This material is based on a joint work with Bruno Da Ré, Emmanuel Chemla and Paul Egré.

November 10th, 2022, Michela Ippolito

The-hell questions

This is a study of expletivethe-hell interrogatives and it’s part of a larger project whose goal is to understand expletive questions and more generally bias in constituent question. We start by discussing counterexamples to two common generalizations: (i) that the-hell questions are aggressively non-D-linked, and (ii) that the-hell questions are always incompatible withwhich-type wh-operators. We show that previous accounts based on the idea that either expletive the-hell triggers adomain extension (e.g. den Dikken & Giannakidou 2002) or a domain restriction eliminating all familiar entities entities (e.g. Martin 2020) are challenged by these counterexamples. We propose that the-hell questions are contextually restricted just like canonical constituent questions and propose that the-hell questions carry a conventional implicature according to which every proposition in the denotation of the question is subjectively less likely (to the speaker) than its negation. We account for both non-D-linked and D-linked occurrences of the-hell questions; for the fact that the-hell questions are restricted to constituent questions; and for the intuition that, in uttering a the-hell question, the speaker believes that no given alternative is viable. Finally, we show that our account of the-hell in questions can be extended to occurrences of expletive the-hell in declaratives.

October, 27th, 2022, Jakob Süskind

Approximate truth and questions

Most people have the clear intuition that some falsehoods are "less false", or "closer to truth", than some other falsehoods. In everyday conversation we often do not hold speakers to the high standard of perfect truth: it is usually sufficient for our claims to be close enough to truth. In the philosophy of science and formal epistemology the notion of approximate truth, or verisimilitude, is of interest as well. It is an important open question whether the intuitive notion of "being closer to truth" actually corresponds to some objective relation between propositions. In this talk, I will present a possible way of tackling the philosophical problem of verisimilitude by making use of the notion of questions: A proposition, I will argue, is the closer to truth the more true answers (and the fewer false answers) it gives to a contextually relevant set of questions. I will try to show how this account solves a number of problems intensely discussed in the philosophical literature on verisimilitude.

October 13th, 2022, James Hampton and Paul Egré

Conjunctions of vague adjectives and their degrees. Joint work with Elisa Scerrati, Gabija Urbonaviciute and Steven Verheyen

In this paper we test an hypothesis put forward by S. Schiffer (2003) regarding the degree to which conjunctive sentences involving distinct vague adjectives, such as "this coat is long and blue", are accepted when the coat is borderline long and borderline blue. Schiffer argued that assuming "this coat is long" and "this coat is blue" each are accepted to degree .5, then normatively "this coat is long and blue" should also be accepted to degree .5. More generally, when P and Q are two vague adjectives expressing probabilistically independent properties, the degree of acceptance of "x is P and Q" ought to be, according to Schiffer, expressed by the min rule of fuzzy logic, instead of the product rule of probability theory. We tested this hypothesis in two phases: first in a between-subject design, and then in a within-subject design. The results of our experiments suggest that the product rule is in fact a better fit than the min rule for such conjunctions. In this talk we will present our results and discuss them in light of the broader distinction established by Schiffer between vagueness-related uncertainty and factual-uncertainty. We also relate the results to earlier findings by Hampton 1988 regarding the combination of vague properties.

October, 6th, 2022, Keny Chatain

Local Pragmatics Redux: Presupposition Accommodation and Non-Redundancy Without Covert Operators

On several occasions in linguistic research, it has been proposed that operations traditionally considered part of pragmatics (such as scalar implicatures or presupposition accommodation) are in fact computed in the semantics by syntactically represented covert operators. The motivation typically comes from observations that these seemingly pragmatic operations are sometimes performed "locally", i.e. in the scope of elements internal to the sentence. Three cases in point: the operator O, which syntacticizes the Gricean reasoning (Chierchia, Fox and Spector, 2012), the operator A, which syntacticizes the process of accommodation (Beaver, 2001, a.o.), and, more recently, the operator R which syntacticizes non-redundancy conditions (Blumberg & Goldstein, 2021). In this talk, we focus on R and A and try to answer the following question: are these operators syntactically real? We use ellipsis and other tests to argue for the negative. We then sketch an entirely pragmatic theory of how accommodation and non-redundancy theories can in some cases be computed locally. Overall, this work aims to raise more general questions: which pragmatic operators are syntactically real, which ones are not ? What does this tell us about the architecture of the semantics/pragmatics interface?

July 21st, 2022, Danny Fox & Moysh Bar-Lev

On fatal competition and the nature of distributive inferences

Denić (2021) observes that the availability of distributive inferences — for sentences with disjunction embedded in the scope of universal quantification — depends on the size of the domain quantified over as it relates to the number of disjuncts. Based on her observations, she argues that probabalistic considerations play a role in the computation of implicatures (likelihood of alternatives determining whether or not they enter into the computation). In this talk we explore a different possibility. We argue for a modification of Denić’s generalization, and provide an explanation that is blind to probabilities. Our proposal is based on the observation that when the domain size is no larger than the number of disjuncts, universal and existential alternatives are equivalent if distributive inferences are obtained. We derive the generalization from a general ban on `fatal competition’: existential and universal sentences are more generally banned when they are equivalent to each other (e.g., # some/all Italians come from a warm country; Magri 2009 and Spector 2014).

July 19th, 2022, Jonah Katz

Gestalt grouping in music and language

In this talk, I review evidence concerning the nature of grouping or phrasing in music and language. The two domains both involve correspondence between auditory discontinuities on the one hand and syntactic, interpretive, or informational constituents on the other, reflecting the Gestalt principles of proximity and similarity. Both domains involve nested, headed, hierarchical constituents. There are also obvious differences between musical and linguistic grouping. Grappling with those differences requires one to think in detail about modularity, information flow, and the functional nature of cognitive domains. I conclude that deep structural and functional similarities exist between musical and linguistic grouping at the computational level, but recovering these structures in real-time at the algorithmic level is necessarily more complex for language. The attempt to compare music and language in this regard illuminates a number of meta-theoretical issues that arise in cross-domain cognitive science.

July 7th, 2022, Keny Chatain

Implicature projection

In recent work, Bassi et al. (2021) amend the grammatical theory of scalar implicatures of e.g. Fox (2007), by proposing that the exhaustivity operator generates presuppositions, rather than at-issue content. In this work in progress, I discuss the consequences of taking a more drastic step: dispense with the operator altogether, and treat all scalar items as introducing projective content (in a way reminiscent of Chierchia (2004)). In this system, interactions between multiple scalar items is always the result of projection and not due to the internal mechanism of an exhaustivity operator.

The main reason to explore this avenue is theoretical: the resulting system is simple. Mascarenhas (2014) shows that in traditional approaches to scalar implicatures (whether Neo-Gricean or grammatical), the resources required for scalar implicature calculation grows (super-)exponentially with the size of the sentence. This observation seems at odds with the rapid on-line derivation of scalar implicature. The projective system I study, on the other hand, only scales linearly.

I show some empirical upsides and downsides of the implicature projection system. On the one hand, it captures a large number of traditionally discussed implicatures, as well as some tricky cases of indirect implicatures, which don't seem derivable otherwise. On the other hand, it is structurally unable to account for various cases of "implicature at a distance": distributive implicatures, Free Choice, implicature of a scalar item in the restrictor of a quantifier. I will show reasons to not give up hope and collect feedback on the best route forward.

June 30th, 2022, Janek Guerrini

Genericity in similarity

In this talk, I offer an account of similarity constructions involving 'like', such as 'be like' and 'look like'. I argue that these constructions have two key properties. (1) The first is that similarity predication amounts to predication of overlap of salient properties: I analyse 'John is like Mary' as 'John shares relevant properties with Mary'. This is motivated by the fact that there seem to be grammatical devices that single out precisely what properties are relevant, e.g. 'With respect to personality, she’s just like her father'. It also corresponds to main accounts of similarity in psychology (Tversky 1977, a.o.), which view objects as sets of features and similarity as a set-theoretical relation between salient subsets of object features.  (2) The second key feature of similarity talk is, I argue, that it involves inherent generic quantification. This explains a range of data: first, it accounts for the reading of indefinites embedded in ‘like' Prepositional Phrases: 'John looks like a lawyer' is almost equivalent to 'John looks like a typical lawyer’. Second, it accounts for narrow-scope and almost conjunctive readings of disjunction in the scope of 'like': 'Mary looks like a lawyer or a judge' is almost equivalent (on its most accessible reading) to 'Mary looks like a lawyer and Mary looks like a judge'.

June 23rd, 2022, Jon Rawski

Generative Capacity & the Linguistic Swampland

A central concern of linguistic theory is determining necessary and sufficient conditions on the generative capacity of the grammars/functions which account for linguistic phenomena.  

Prior work shows that while the landscape of possible functions is vast, the range of "things linguistic" is tightly constrained.

In this talk I will review some recent mathematical work describing (1) conditions on the generative capacity of phenomena independent of a particular framework, using techniques from mathematical logic, and (2) classifying different linguistic theories with respect to these mathematical conditions.

The key lesson is that many linguistic theories lie in what I call the "swampland", a vast area where the capacity of the theories does not obey these constraints, even if the theories are themselves internally consistent-looking.

This talk will use phenomena from our recent work in morphology, mostly reduplication (Dolatian et al 2021, in review), with some helpful grounding using syntax & semantics.

June 16th, 2022, Michael Franke

Helpfulness of answers and goal-signaling questions

Relevance is a central notion in the study of communication, but it is multifaceted, if not elusive and therefore hard to approach formally. This talk tries to get closer to a formalization of useful notions of relevance by looking at experimental data based on which different information-theoretic concepts of relevance of an answer can be compared. It then introduces a probabilistic model of question and answer choice which pivots around an action-based notion of relevance of information and which predicts how helpfulness of an answer is informed by the goal-signaling quality of relevant questions. 

June, 9th, 2022, Maël Leroux (Animal linguistics communication)

Call combinations in chimpanzees?

Syntax, i.e. the ability to combine meaning-bearing units into larger structures, is considered thekey feature distinguishing human language from other non-human animal communication systems. However,evidenceforcombinatorial capacities in non-human primates has challengedthisassumption and suggests syntax might have its origins deeply rooted in the primate lineage. By investigating such capacities in chimpanzees, our closest-living relatives, Iaim to elucidate whether syntax is a de novo evolved trait inhumans or whether it is evolutionarily more ancient.

Taking a whole repertoire approach,I firstly show chimpanzees produce distinct non-random call combinations, primarilyin socially challenging situations. Secondly, using behavioural data, I explore the potential function of the “pant-hoot + food call” combination. Chimpanzees produce this structure specifically whilefeedinginlarge trees,whenhigh-rankedindividuals join.By transferring information on the identity of the signaller and the presence of food, this combination could serve two functions: i) attract important social partners to a valuable food source and/or ii)prevent potential future aggression. Finally, using an experimental approach, I provide strong evidence that the“alarm-hoo+ waa-bark” combinationis meaningful to receivers. Chimpanzeesproduce this combination when encountering a snake while being isolated from the group, potentially to recruit other group members in dangerous situations. Furthermore, receivers extract this structure’s meaning in a compositional way (i.e. “alarm-hoo” + “waa-bark” = “danger”+ “recruitment”).

Together,these data provide key support that combinatoriality is a core component of chimpanzee communication and that these sequences are communicatively relevant and meaningful to receivers. Moreover, findings on combinatorial abilities in our closest-living relatives fill an important missing piece of the evolutionary puzzle of syntax and support the hypothesis that syntax emerged gradually from our common ancestors with our primate cousins.

June, 2nd, 2022, Nina Haslinger

Vague and non-vague non-maximal predication: Revisiting the plural Sorites paradox

This talk revisits the relation between imprecision and vagueness. The imprecision phenomenon I will focus on is non-maximality in plural predication: While the default construal of (1a) is one that makes it truth-conditionally equivalent to (1b), there are contexts in which (1a) can be true even if some of the contextually relevant switches are on.

(1)

a. The switches are off.

b. All the switches are off.

My aim is to bring together two (prima facie) conflicting lines of research: On the one hand, Burnett (2017) proposes a unified approach to non-maximality and vague predication within the "strict/tolerant" framework (Cobreros et al. 2012b), motivated by the observation that non-maximal plural predications can give rise to the Sorites paradox. On the other hand, non-maximal interpretations often have precise truth conditions, and non-maximality does not pass certain tests for vagueness such as the acceptability of "pseudo-contradictions" like (2), which has led to the development of the QUD-based approach to non-maximality, on which vagueness and non-maximality are unrelated (Malamud 2012, Križ 2015, Križ & Spector 2021, Bar-Lev 2021, Feinmann 2020 a.o.). In this framework, the QUD directly influences the truth and falsity conditions of plural sentences; essentially, non-maximal interpretations are available if the difference between maximal and non-maximal scenarios does not matter for the purposes of the QUD. This makes largely correct predictions about the contextual constraints on non-maximal interpretations of plural sentences, but predicts the truth conditions of these sentences to be precise in every context.

(2)

a. #The switches are off and they are not off.

b. John is tall and he is not tall.

This tension in the literature can be resolved by looking more closely at the contexts in which plural predications give rise to the Sorites paradox: These are typically contexts in which there is no clear-cut contextual goal or binary QUD, but rather a relatively large set of potential outcomes, some of which differ only negligibly in their consequences. Essentially, the contexts giving rise to vague non-maximal construals are contexts in which it is underdetermined what exactly is at issue (cf. Graff 2000), because there are several similarly "good" choices for the QUD. In contrast, the QUD-based non-maximality literature has mostly focused on contexts in which there is a well-controlled, binary issue.

Based on on this characterization of the contexts that permit vague non-maximal interpretations, I propose an analysis that combines the QUD-based framework with a super-/subvaluationist version of the strict/tolerant idea, as proposed in Cobreros et al. (2012a).  I suggest that the QUD-based theory of Križ & Spector (2021) should be thought of as a theory of *strict* truth conditions relative to a question Q. The *tolerant* truth conditions relative to Q are then derived by considering certain subquestions of Q that unify some of the partition cells of Q, while not diverging too much from Q. A plural sentence counts as tolerantly true iff it is true relative to at least one such subquestion. This divergence between strict and tolerant truth conditions makes it possible to apply the account of the Sorites paradox in Cobreros et al. (2012a).

On this proposal, the QUD-dependence of imprecise expressions is directly tied to their potential for vagueness, and also accounts for the optionality of vagueness with such expressions. The proposal also predicts an indirect connection between vagueness in plural predication and homogeneity: Vague non-maximal interpretations always give rise to homogeneity effects, while this is not necessarily the case for non-vague non-maximal interpretations. Further, it avoids some empirical problems of the earlier strict/tolerant approach of Burnett (2017) while preserving its insights.

Time permitting, I will conclude by discussing the status of "pseudo-contradictions" as in (2) — which are argued to be a test for context-dependency that feeds semantic composition, rather than a test for vagueness (cf. also Feinmann 2020) — and (building on joint work with Mathieu Paillé) some potential extensions of the proposal to imprecise construals of total/partial predicate pairs like *open/closed*.

May, 25th, 2022, Naomi Francis

Even in presupposition denials

This talk explores a puzzle about even and its crosslinguistic kin. These scalar additive operators can be used in negative sentences that deny presuppositions, but not in their positive counterparts, as illustrated in (1) for English.

May, 19th, 2022, Milica Denić

The trade-off between storage and use complexity in natural languages

Human languages vary in terms of which meanings they lexicalize, but there are important constraints on this variation. It has been argued that languages are under pressure to be simple (e.g., to learn or to store in memory) and to be informative (i.e., to allow for a precise communication), and that these two pressures are driving forces in determining which meanings get lexicalized (Kemp and Regier, 2012 and much subsequent work). I will argue that the way informativeness is operationalized in that line of work is problematic in that it doesn't take into account the fact that we can often precisely express a meaning which is not lexicalized in our language by means of a morphosyntactically complex phrase. I will then propose a different perspective on what pressures determine which meanings get lexicalized: the pressure to lexicalize as few concepts as possible (storage complexity) and the pressure to produce as morphosyntactically simple utterances as possible (use complexity). I will present a case study showing that numeral systems in natural languages optimize the trade-off between storage and use complexity, and discuss how these ideas relate to existing proposals on pressures shaping language.

May, 12th, 2022, Eszter Ronaï

Quantifying semantic and pragmatic effects on scalar diversity

Scalar inference (SI), the process by which we systematically infer meanings stronger than what was explicitly said, has long been a central topic of investigation in semantics-pragmatics. A recent experimental finding that has generated interest is scalar diversity: that the robustness of SI calculation varies across lexical scales (i.a., van Tiel, et al., 2016). For instance, the SI in (1) is much more likely to arise than the one in (2).

(1)  Mary ate some of the cookies. —> SI: Mary didn’t eat all of the cookies.

(2)  The movie is good. —> SI: The movie is not excellent.

In this talk, I will first present a proposal to more rigorously quantify the observed variation across scales, using relative entropy. I will then turn to the question of how factors independent of scalar diversity can make inference calculation both more likely and more uniform. Specifically, I will look at the role of discourse context, as in (3), and the role of semantic exhaustification with the focus particle only, as in (4).

(3)  A: Did Mary eat all of the cookies?

      B: She ate some of the cookies.

(4)  Mary ate only some of the cookies.

I will present experimental data showing that a supportive discourse context and overt exhaustification with only both increase inference rates and they both reduce variation across scales—with the effect of only being stronger. However, there still remains a lot of scalar diversity; only when we combine context with semantic exhaustification do we find uniformity across lexical scales.

April, 21st, 2022, Paul Egré & Benjamin Icard

VAGO: an online tool to measure vagueness and subjectivity in discourse

VAGO is an annotated lexical database used to provide a measure of vagueness and subjectivity in discourse, which we designed and for which we are developing a testable online tool with Ghislain Atemezing (MONDECA). It rests on a four-fold typology of vague expressions, distinguishing generality, approximation, one-dimensional and multi-dimensional vagueness. Currently, the database consists for the most part of adjectives, and it consists of a limited lexicon for English and for French (of about 1500 words in each language). We use the typology to basically label expressions as markers of subjectivity or objectivity. We will explain the motivations for the typology, and the way it can be used to help the evaluation of opinion-like vs fact-like statements. We will explain how it can be used for the detection of fake news articles, in particular combined with deep learning methods developed by other partners. We will also highlight several limits of the framework, and discuss directions for future improvement.

April, 14th, 2022, Jean-Baptiste André

Biological contracts and the evolution of morality

Our goal in this paper is to account for both the existence and the fine-grained properties of human morality with an evolutionary perspective. Our approach is based on the premise that buman beings are under selection to appear as good cooperative investments. Hence they face a trade-off between maximizing the immediate gains of each social interaction, and maximizing its long-term reputational effects. In a simple 2-player model, we show that this trade-off leads to the generalized Nash bargaining solution being played at evolutionary equilibrium, and we infer from this result the general theoretical proposition that morality is a domain-general calculator of this bargaining solution. We then proceed to describe the generic predictions made by the theory: (i) everyone in a social interaction deserves to receive a net benefit, (ii) people ought to act in ways that would maximize social welfare if everyone was acting in the same way, (iii) all domains of social behavior can be moralized, (iv) moral duties can seem both principled and non-contractual, and (v) morality shall depend on the context. Next, we apply the theory to some of the main areas of social life and show that it allows to explain, with a single logic, the entire set of what are generally considered to be different moral domains. Lastly, we discuss the relationship between this theory and other evolutionary theories of morality and cooperation.

April, 7th, 2022, Maxime Cauté (Animal Linguistic)

Cross-modal compositionality in chimpanzees: A tentative compositional semantics for face-gesture interactions

A recent article from Oña et al. (2019) observed face-gesture combinations in chimpanzees. They noted that the response to a gesture differed when a face was added, yet the effect of a face differed across gestures. They infered that these face-gesture combination was neither holistic nor compositional, but a mixture of both: a property they brand componentiality. We will challenge this view by a) offering a theoretical compositional account of such inconsistency; and b) by providing a tentative semantics for the actual data gathered by Oña et al., in line with the existing literature.

March, 24th, 2022, Isidora Stojanovic

Exploring Valence in Judgments of Taste (based on joint work with Elsi Kaiser, USC)

Judgments of taste are often assumed to be evaluative, in the sense of expressing either a positive or a negative assessment of the object under evaluation. However, the evaluative character of predicates of personal taste (PPTs) has received relatively little attention from a semantic point of view. Our aim is to fill out this lacuna. We show that PPTs do not divide neatly into positively and negatively valenced terms. Instead, we suggest that many PPTs, such as ‘surprising’ and 'intense', are neutral: they are underspecified for their valence and, depending on the context, can give rise to a positive, a negative, or an ambivalent evaluation. We investigate how such neutral PPTs differ from evaluative PPTs, and how they differ from certain other terms that are neither positive nor negative, such as ‘average’. We use a two-pronged approach. First, we propose two novel linguistic tests that serve as diagnostics to distinguish the class of neutral PPTs from valenced PPTs, and we use corpus examples to corroborate the tests. Second, we use information from pre-existing psychological norms of valence to further explore the class of adjectives that we hypothesize are neutral.

March, 17th, 2022, Nicolas Guérin

What semantics for deep learning systems, what do they understand about content and logical word

Deep learning systems achieve amazing successes in a variety of language tasks (e.g., translation, natural language inference, sentiment analysis, …). They are typically trained on a simple task, roughly: predict the next word in a sentence. This is a very syntactic task, if we understand syntax as the constraints on the (joint) distribution of words in sentences. In learning to master this syntactic task, what semantic knowledge do these systems capture?

We prompt semantic knowledge with internal consistency tests, or proof-theory tests, that is with no need for ‘grounding’ in the real world. For content words, we ask whether the systems recognize that “bird” is a good continuation in “A seagull is a type of  ____”. For function words, we focus on words such as “therefore” or “because”, which meanings relate directly to entailment, a foundational aspect of what the whole of semantics is about. Do deep learning systems recognize that “therefore” is appropriate in “This is a bird, <therefore> this is a seagull”?

March, 10th, 2022, Florian Schwarz & Alexandros Kalomoiros

Presupposition Projection and Linear Order: Reconciling an Explanatory Mechanism and Variation across Connectives

The role of linear order for presupposition projection is a long-standing matter of theoretical controversy and empirical confusion. On the one hand, in the Stalnakerian tradition, in particular in the influential reframing by Schlenker (2009), there’s a natural association of the `left-to-right’ unfolding of the linguistic signal and the gradual update of the contexts relative to which subsequent expressions are interpreted. On the other hand, connectives seem to display varying behavior in terms of whether later material can affect presuppositions of earlier expressions. In particular, conjunction has often been taken to be asymmetric with regards to projection, only allowing `left-to-right’ filtering (e.g., with the first conjunct supporting a presupposition in the second conjunct), whereas disjunction seems to also allow the reverse `right-to-left’ filtering (as in Partee’s `bathroom sentences’). But the validity of these commonly reported generalizations has been challenged, too, e.g., by arguing that the apparent asymmetry of conjunction is due to certain confounds (Rothschild, 2011), or by appealing to other theoretical mechanisms, such as local accommodation, to explain away `right-to-left’ filtering with disjunction (Hirsch & Hackl, 2014). The resolution of these issues is of central importance for theories of presupposition projection, as they bear on the key question of whether or not a general and explanatory account of the role of linear order for projection along the lines of Schlenker (2009) can be maintained, or whether variation in order effects across connectives undermines this enterprise.  

We present a series of experimental investigations of projection from both conjunction and disjunction. Our results reported in Mandelkern et al. (2020) show that projection from conjunction is genuinely asymmetric (i.e., limited to left-to-right filtering), and ongoing work with Alexandros Kalomoiros establishes that projection from disjunction is genuinely different from conjunction in this regard (by also allowing genuine right-to-left filtering). These results challenge a general account of projection, such as Schlenker’s Local Context approach, which predicts uniform effects of linear order on projection across connectives.

In the final part of the talk, we outline the Local Symmetry account currently being developed by Alexandros Kalomoiros, which aims to reconcile the generality and explanatory appeal of Schlenker’s approach with variation across connectives by letting their truth-conditional profile impact the effect of linear order in the newly formulated projection mechanism. In addition to showing how Limited Symmetry accounts for differences between simple conjunctions and conjunctions, we spell out its further predictions for various construction and report some initial data testing these.

February, 17th, 2022, WooJin Chung

Local context is calculated domain by domain, for each maximal projection

Theories of local contexts that either rely on strict linear order (Schlenker 2009) or hier- archy (Ingason 2016, Romoli and Mandelkern 2017) face a number of challenges, most notably in nominal modification (Schlenker 2020, Anvari and Blumberg 2021) but also, depending on the theory, in accounting for attitude verbs and conjunction in head-final languages. To resolve such issues, I propose a hybrid theory which takes into account both linear order and syntactic structure: local context is calculated domain by domain, and for each maximal projection.

February, 3rd, 2022, Pritty Patel-Grosz

The search for universal great ape gestures in chimpanzees, bonobos, and humans

We explore the question of whether interactive, discourse-managing gestures in humans can be traced to directive gestures in chimpanzees as part of a potentially universal gestural inventory shared throughout the family of great apes (humans, chimpanzees, bonobos, gorillas and orang-utans). We carried out two exploratory experiments (n=300 each) in which we tested (i) whether human participants would infer the same meanings that are attested in chimpanzees for chimpanzee gestures performed by a human gesturer; (ii) whether human participants would infer interactive gestures’ meanings that derive from the chimpanzee meanings by virtue of a conduit metaphor (consisting of a transfer from object-manipulating meanings to discourse-managing meanings). Our experiment is embedded in the outline of a formal semantic implementation that treats human interactive gestures as discourse-managing devices, such as, for example, denial-related operators.

January, 20th, 2022, Moysh Bar-Lev

Language stability and the typology of logical connectives (joint work with Roni Katzir)

Most accounts of the typology of natural language connectives stipulate that “and” and “or” are inherently simpler than other connectives, which comes close to assuming what needs to be explained. In recent work, Enguehard & Spector (2021; E&S) propose to replace this stipulation with assumptions about the prior probabilities of states. In this work in progress we follow E&S in exploring the possibility that the typological pattern can be explained by considerations having to do with prior probabilities of states, but unlike them we do not make any specific assumptions about priors. Instead, we consider how different state priors may affect a rational speaker’s message choices given a model of communication (the RSA model). We observe that attested languages are stable languages, i.e., languages in which the optimal message for a speaker to choose when they want to convey a particular state they are in is not affected by which states they take to be more likely and which ones less. We discuss how this observation about stability can be utilized to explain the typology, and what implications upholding this explanation has on the role of probabilities in communication, especially in light of problems raised by Fox & Katzir (2021) for communication models which are affected by prior probabilities.

December, 16th, 2021, Diego Feinmann

Contradictoriness

Why does a sentence such as ‘John was killed but didn’t die’ feel contradictory? The standard response to this question is familiar: ‘John was killed but didn’t die’ feels contradictory because it is necessary false (i.e. it is false in every possible world). This talk is divided into three parts: in the first part of the talk, I show that the condition of being false in every possible world isn’t a predictor of contradictoriness; in the second part, I put forth an account of contradictoriness and discuss its predictions; in the third part, I make the case that redundancy is the other side of the coin of the phenomenon of contradictoriness and suggest a path to theoretical unification.

December, 9th, 2021, Isabelle Dautriche & Emmanuel Chemla (animal communication)

Language-like operations in baboons

We identify elementary skills in the language abilities of humans, and trace them in the baboons in Marseille. We report three studies suggesting that baboons (i) like to manipulate 'concepts' of the same shape as ours, (ii) can respond to negation-like operators <t,t>, (iii) they report responses in a "compositional" manner.

December, 3rd, 2021, Roberta D'Alessandro

Heritage language syntax: a view from microcontact

This talk will discuss some of the results of the ERC project Microcontact, involving several Italo-Romance languages in contact with each other. First, some data will be discussed showing that the general laws that are considered valid in driving change in contact do not hold for microcontact. Then, some speculations will be proposedon the role of verb movement to enlarge domains of computation. It will be shown that in heritage varieties where verb movement is lost or weakened, the computational domain is restricted. This situation will be compared to that of the emergence of Old Italian cliticization, which will also be argued to be a domain-delimiting mechanism.

December, 2nd, 2021, Roberta D'Alessandro

Understandinglanguage change – directionality, simplification, or neither?

One of the questions regarding language change is whether the change that happens as a result of language contact is of the same nature as that which happens indiachrony. According to some scholars, change in contact is just “accelerated” diachronic change, and there is fundamentally no difference between the two.The keyword is “simplification”: change almost always happens to simplify the system. Witness of this are the urban contexts in which many languages are spoken by a large number of non-native speakers, where some sorts of simplifiedversions of the “official” standard language (Trudgill 2011).

In this talk, I will explore the extent to which a rather large sample of Romance languages in microcontact (i.e. contact among languages that are typologically identical and only present very limited grammatical variation) confirms this claim. After a general introduction to the methodology of microcontact linguistic investigation, some data generalizations will be considered regarding deixis, auxiliary selection, differential object marking, and null subjects in contact. It will be shown that

-pronouns in (micro)contact (Terenghi 2021) confirm the claim that change in contact and diachronic change move along the same direction, and pronouns do not ever evolve in a direction different from that in which they have been shown to evolve independently.

-auxiliary selection in contact show that change moves along a “simplification”path, which is very similar to that described in grammaticalization studies. However, if the simplification is confirmed, the choice of the auxiliary is different: while Italo-Romance heritage in America tend to extend HAVE, the same varieties in Italy have witnessed the expansion of BE.

-Differential Object Marking in contact shows a tendency which is different from that of the diachronic development, in that DOM in contact tends to disappear (which is a sort of simplification). However, microcontact, as well as extreme “macro-contact” such as that which is found in creoles mirrors the diachronic development (i.e. emergence of DOM, complexification).

Based on these generalizations, I will argue that not all contact situations can be considered as equivalent, but that both diachronic change and change in contact are very often driven by general cognitive principles (such as linking, or topicality) to which both 1st language learners and speakers of contact varieties resort to find some regularity in the messy language data. In substance, the questions about simplification or directionality of change are ill-posed; rather, if speakers have no clear perception of the locus of variation, they will adapt “third-factor”/cognitive strategies to resolve the conflict between the grammars in contact. If the locus of variation is salient and perceptible instead, change will follow a grammatical path to conflict resolution. Hence, there are some common strategies to the resolution of change in contact, but they are not simplification.

November, 25th, 2021, Morgan Moyer

Understanding the semantics and pragmatics of (non-)exhaustivity in wh-questions with corpus and experimental methods

Asking and answering questions is a staple of human communication. In order to answer a question effectively, a hearer must interpret the speaker's intention given the specific question asked. Wh-questions like Where can I get coffee? are underspecified for (non-)exhaustivity, i.e., how many answers must be provided to resolve the speaker's goal. Intuitions from the semantic literature report that questions are generally exhaustive, and non-exhaustive only in the context of specific linguistic factors (e.g., the modal 'can', certain wh-words). To test these assumptions, we collected question paraphrase ratings for naturally occurring root questions in variable linguistic contexts. In contrast to previous claims, we find that questions are not biased for exhaustivity. However, other prior observations are supported by the data. We argue that a full account of the observed distribution of meanings must integrate discourse factors like the hearer's estimate of the speaker's goal, alongside (or subsuming the effect of) linguistic cues.

November, 18th, 2021, Itai Bassi

On Presuppositional Exhaustification and the semantic-pragmatic status of Scalar Implicatures

Grammatical Theories of Scalar Implicatures (SIs), on which (some) SIs are generated as part of the procedure for compositionally interpreting logical forms, are usually cashed out using a syntactic operator exh which contributes a SI enrichment at the at-issue level of meaning. Though successful in explaining various phenomena that are problematic for non-grammatical approaches to SIs, exh-based theories have sometimes been criticized as not providing an adequate account of: (i) the non-at-issue status of implicatures (Russell 2012), and (ii) the restricted distribution (or even absence) of SIs in a number of embedded positions. In this talk, based on recent and ongoing work with Guillermo del Pinal and Uli Sauerland, I present a new grammatical theory which aims to address these criticisms. The heart of the proposal is that exh is to be replaced by pex, an operator which contributes SI at a non-at-issue, presuppositional component of meaning. I will try to show that pex, together with independently motivated generalizations about the behavior of presuppositional content and about the role of Relevance in SI computation, provides an improved account of (i) the pragmatic status of SIs, (ii) the puzzling behavior of SIs projected from under negative factive predicates (Gajewski & Sharvit 2012;  Spector & Sudo 2017), and (iii) the well-known restricted distribution of SIs under negation. If there is enough time, I will present an extension of the pex-based theory to deal with Free Choice (FC) inferences, and show how it explains some recent data about projection and filtering free choice (Romoli and Santorio 2019, Marti and Romoli 2021).

November, 4th, 2021, Alexandre Cremers

Ignorance inferences and roundness effects with modified numerals

Modified numerals like "at least four" or "more than three" are notorious for a number of puzzles they pose to the theory of implicatures. While 'more than 5' and 'at least 6' should intuitively be synonymous, 'at least' obligatorily conveys ignorance (hence the weirdness of "I have at least three kids") but 'more than' doesn't (in a context with a salient threshold, "I have more than three kids" is acceptable). Conversely, 'at least' can combine with any numeral, but 'more than' shows a strong preference for round or contextually salient numerals ("Mary can drink, she's more than 23" is degraded when the relevant threshold is 18). I will first present experimental results that shed new light on seemingly conflicting results from previous empirical studies on the ignorance implicatures of 'at least' and 'more than', before presenting a new account of modified numerals based on a very simple semantics and a new pragmatics formalized in Optimality Theory. The basic model captures all standard properties of modified numerals with minimal assumptions: obligatory ignorance with 'at least' but not with 'more than', QUD effects, roundness sensitivity of 'more than' but not 'at least', and basic polarity effects.

October, 28th, 2021, Keny Chatain

Homogeneity properties of cumulative readings of quantifiers

Cumulative readings of singular quantifiers (Schein 1996), illustrated in (1a), have long been seen as challenging to account for from the perspective of standard compositional assumptions.

(1)

a. The children ate every cake.

b. The children didn't eat every cake.

In this talk, I argue that by studying the truth-conditions of the negative (1b), we can gain insights on how sentences like (1a) and (1b) might be composed. Specifically, taking inspiration from Bar-Lev (2018), I argue that we ought to distinguish between an unstrengthened meaning and a strengthened meaning for sentences like (1). I show that the unstrengthened meaning of such sentences in fact obeys standard rules of composition. The strengthened meaning, derived from the unstregthened meaning by addition of an implicature, obscures this underlying simplicity. This analysis predicts intricate cases, including Schein's famous video-game examples, as well as asymmetries in the availability of cumulative readings of quantifiers.

October, 21st, 2021, Jeremy Kuhn, David Nicolas, Brian Buccola

Groups versus covers revisited: Structured pluralities and symmetric readings 

A number of natural language constructions seem to provide access to structured pluralities — that is, pluralities of pluralities. A body of semantic work has debated how to model this additional structure and the extent to which it depends on pragmatics. In this talk, after controlling for the distinction between ambiguity and underspecification, we present new data showing that structured pluralities are sometimes but not always available, depending on the form of the plural noun phrase used. We show that these results challenge two longstanding theories of plurality. We sketch two different ways to account for these data and describe some of the diverging predictions they make.

October, 7th, 2021, Alexandre Cremers

Testing probabilistic models of exhaustivity 

(joint work with Ethan Wilcox and Benjamin)

The RSA has successfully captured a number of linguistic phenomena at the semantics/pragmatics interface, and offers a new formalization of Grice's theory of implicatures, in particular the fact that "some" is often understood to imply "not all". However, the model also makes the counter-intuitive prediction that when the prior for an all-situation is high, "some" could in fact convey "all" rather than "not all". Degen et al. (2015) confirmed our intuition that this prediction was not borne out, and proposed a variant of the RSA which addresses the issue by downplaying the importance of priors in the model (de facto departing from the assumption that listeners are Bayesian). In this talk I will discuss another kind of implicatures: exhaustivity inferences. If you ask Sue what she had for lunch and she answers "A sandwich", you will typically infer that she had a sandwich and nothing else (i.e., that her answer is exhaustive). This is analogous to the some/all case except for the costs of the various messages. I will present a new study which tested the effects of prior beliefs on both production and comprehension in this situation, improving on previous empirical work. The results confirm that participants' behavior is in clear contradiction with the predictions of RSA, for both production and comprehension. The alternative model proposed by Degen et al. captures the data, but it must assume unrealistic values for some of its parameters. A model built on Spector's (2017) model for homogeneity offers a better fit of the data with acceptable parameter values. The crucial feature needed to capture the data seems to be the ability to break the symmetry between stronger alternatives in the grammar, rather than relying on cost and informativity alone.

July, 15th, 2021, Patrick Elliott

Externally-Dynamic Dynamic Semantics

In this talk I revisit an empirical domain foundational to dynamic approaches to meaning: anaphora to singular indefinites in complex sentences, and in discourse. First-generation dynamic theories (Heim 1982, Kamp 1981, Groenendijk & Stokhof 1991) use a set of canonical generalizations to motivate grammars according to which the flow of referential information is pre-compiled into lexical items. We'll raise a number of issues for the facts taken for granted by first-generation dynamic theories, motivating an alternative, more modular approach based on the following principles: (i) logical operators are simply truth-functional operators, and therefore blind to referential information (ii) a single "switch" is responsible for referential information flowing uniformly throughout the grammar. We'll formalize these ideas in a compositional fragment using an applicative functor (Kobele 2018, Charlow 2019a, 2019b). The result will be a more principled theory of anaphora, that is nevertheless rather permissive; we'll see that restrictiveness (re-)emerges once we take independently motivated pragmatic principles into account.

July, 1st, 2021, Léo Migotti & Eleonore Scholler

Music semantics within auditory cognition

For a few years now, musichas been argued to have a semantics (Schlenker 2017, 2019, 2021), i.e. musiccan evoke a music-external reality in listeners. We present an experimental paradigm that aims at testing what musical items are responsible for these semantic associations, how these items are systematically interpreted in non-musical terms, and where these associations come from. We hypothesize that subjects are indeed able to make such associations in a productive fashion, and that the cognitive algorithms responsible for these associations are partly grounded in the general, non-musical auditory cognition. We present preliminary results that support these claims and discuss possible theoretical and experimental extensions.

June, 17th, 2021, Mylène Dutour (animal communication)

Interspecific communication in birds

I am a behavioural ecologist interested in how different species use acoustic signals during collective and social behaviours. I defended my PhD at the University of Lyon in 2018 about the acoustic communication systems of the passerine birds during predator mobbing – a behaviour that happens when prey species discover a predator, and rather than flee, they attack the predator and produce mobbing calls until it leaves the area. Specifically, my research focused on how species encode information in their mobbing calls, how conspecifics and heterospecifics respond to mobbing calls, and what are the mechanisms and the extrinsic factors explaining these responses. Currently, I am a postdoctoral fellow at the University of Western Australia working on the positive relationship between cognition and fitness in Western Australian magpies (Cracticus tibicen dorsalis). I am conducting experiments to determine if smarter females are better at protecting their young, or provide to their young better quality food. More recently, I have begun to work on heat stress and interspecific interactions in magpies.

June, 10th, 2021, Benjamin Spector

Anaphora without dynamic semantics - A static trivalent semantics for pronouns and indefinites, combined with Schlenker's transparency theory for presupposition projection

Current theories of presupposition projection fall, by and large, into three categories:

- dynamic theories (Heim 1983, and much subsequent work)

- static trivalent approaches (Peters 1997, Beaver & Krahmer 2001, a.o)

- static bivalent approach supplemented with the Transparency Principle, according to which presupposed content must be redundant in the syntactic environment in which the presupposition trigger appears (Schlenker 2007, 2008), or Chemla's Similarity Principle (Chemla 2008).

It has been recognized long ago that the phenomenology of presupposition projection is closely related to that of the licensing of anaphora. That is, the conditions under which a non-referential pronoun can be anaphoric to a previously introduced indefinite (as in `Ada bought a pen and she used it to write a letter' ) are very similar to those under which the presupposition triggered by some expression can be `filtered out' by previously introduced content (as in `Ada bought a pen and John knows that she bought a pen').

This striking resemblance is well accounted for in dynamic theories of anaphora (Kamp 1981, Heim 1982, Groenedijk & Stokhof 1991 and much subsequent work) which are conceptually related to dynamic theories of presupposition, or even unified with them (van der Sandt & Geurts 1991, van der Sandt 1992) . But, with the exception of unpublished work by Philippe Schlenker, nobody has yet proposed a non-dynamic theory of anaphora which would use the conceptual ingredients of non-dynamic theories of presupposition, based on either trivalent semantics or the Transparency Principle.

In my talk, I will propose a non-dynamic, static approach to anaphora which combines a trivalent approach with Schlenker's principle of Transparency. I will make use of a key idea recently introduced by Matt Mandelkern (in his manuscript `Witnesses'), according to which an existential statement such as 'Someone_x came' introduces a variable (notated x here), and, relative to a world w and and assignment function g, is undefined, rather than true or false, if it is classically true in w (i.e. if someone came in w) but g(x) does not denote an individual who came in w. Furthermore, I will assume that a sentence such as `She_x came' comes with the presupposition that the variable x is "valued* by the assignment function (which can be partial), a presupposition which is subject to the Transparency principle. Importantly, in the system I will propose, undefinedness will not be equated with Presupposition Failure (e.g., even though 'Someone_x came' can be undefined, it is in fact never a presupposition failure) . Rather, presupposition projection will be governed by the transparency principle: the presupposition `x is valued' should be redundant in the syntactic position in which `She_x came' occurs. While related to Mandelkern's approach, my proposal, unlike his, is fully static at the level of compositional semantics. For convenience, I will present the core ideas applied to the language of predicate logic (into which the relevant English sentences will be translated).

June, 3rd, 2021, Diego Feinmann

On Hyperbole

A hyperbole is an exaggerated statement that isn’t meant literally andis used for effect (typically, to convey affect or make an evaluation of eitherpositive or negative valence). For example, I may utter ‘The water in the bathis boiling’, despite knowing that the water in the bath isn’t boiling, tocommunicate that the water in the bath is too hot—and that I’m frustrated as aresult of this. Since Grice (1975), this phenomenon has been thought to befully pragmatic in nature (cf.Sperber and Wilson 2008; Wilson 2011; Kao, Wu, Bergen and Goodman 2014). In thistalk, I put forward a contrarian idea: hyperbole interpretation is partly asemantic phenomenon (come to the talk if you want to know more).

May, 20th, 2021, Amir Anvari

Content and Redundancy

Much recent work in theory of assertability has focused on conditions under which a sentence is infelicitous due to redundancy (Schlenker 2008; Meyer 2014; Katzir & Singh 2014; Mayr & Romoli 2016; Ciardelli & Roelof- sen 2017; Mandelkern & Romoli 2018; Marty & Romoli 2021). While crisp generalizations pertaining to disjunction in particular have emerged from this work, the theoretical consequences have remained puzzling (Marty & Romoli 2021). With the exception of Ciardelli & Roelofsen (2017) and Sudo (2017), all work in this area has resulted from attempts to maintain a classical semantics and designing a suitable theory of redundancy to capture the relevant generalizations. I will adopt the simplest theory of redundancy available, based on a global anti-redundancy principle, and ask what sort of semantics captures the generalizations in question coupled with this theory of redundancy. In particular, I will show that ‘basic inquisitive semantics’ (Roelofsen 2013, a.o.) provides an elegant solution. Various theoretical consequences pertaining to exhaustification and presupposition projection will be discussed as well as problems with conjunctions with entailing conjuncts and ‘Hurford conditionals’ (Mandelkern & Romoli 2018).

May, 6th, 2021, Can Konuk

Logical reasoning without Intellectualist postulates

Most research on reasoning has relied on an `Intellectualist’ postulate: reasoning is after truth. Logical mistakes are bugs of reasoning which Intellectualist theories should build additional hypotheses to account for. Recently, the Interactionist theory of reasoning proposed a different approach. Reasoning is not for tracking truth but for convincing others with arguments. This converse approach comes with a converse problem : how can humans consistently agree on logical norms and show a preference for logically valid arguments? This porblem is hard to fix within the Interactionist framework without reintroducing Intellectualist assumptions. I argue however that a solution can be offered by cashing out the Interactionist theory with a lower-level theory of reasoning, the Erotetic theory of reasoning (Koralus \& Mascarenhas, 2013). I prove that certain structural properties of the Erotetic system make it so that more logical arguments systematically rely on less informative premises as a coextensive feature. This allows an evaluation function that tracks the informativeness of arguments, in the style of the Interactionist theory to effectively favor logical reasoning in Interactions without being tailored to classical logic. This proof will have an important theoretical implication: we can explain the normative power of classical logic on human reasoning without assuming that the norm of classical logic is implemented anywhere in the human computational system.

April, 29th, 2021, Thomas Brochhage

Consolidating linguistic tendencies at the level of individuals and populations for Zipf's Law of Abbreviation

Communicative pressures can give rise to regular patterns of language use. These patterns, in turn, can come to shape a language’s structure over time. In a recent study, Kanwal et al. (Cognition, 165:45–52, 2017) investigate whether an interaction of such pressures may underlie the cross-linguistic tendency of frequent forms to be shorter. This experiment was conducted among dyads, however, the regularity it seeks to explain is at the level of languages. In this talk, I build on Kanwal et al.'s data to reflect on the tension between individual- and population-level tendencies; on how linguistic change at one level may influence another; and on the work that still needs to be done to link these levels of analysis. Along the way, I will show that a multi-level probabilistic speaker model best characterizes both individual subjects' behavior as well as that of the population as a whole.

April, 15th, 2021, Patrick Grosz

Face Emoji - a Case Study in Expressive Meaning

(joint work with Gabriel Greenberg [UCLA], Elsi Kaiser [USC] and Christian De Leon [UCLA])

This talk argues that face emoji (e.g. the Grinning Face emoji and the Worried Face emoji) are part of a multi-modal discourse and express emotional attitudes. We provide evidence that they comment on the text that they accompany, and that they are more constrained - in semantically interesting ways - than one might initially expect. We propose that they denote functions from an individual x (the attitude holder) and a proposition p (the object of emotion) to sets of situations in which the individual x holds an emotional attitude (e.g. happy or unhappy) about how the proposition p relates to salient contextual factors, such as a background question Q. We argue that this analysis derives a range of interesting constraints, including [i.] positioning constraints of emoji with regards to the text that they accompany, [ii.] apparently direct interactions of emoji with lexical material in the accompanying text (including the scalar operator ‘only’), [iii.] and a range of apparent mixed-emotion uses of face emoji. The talk will conclude with a discussion on how face emoji relate to expressive meaning in natural language (such as the curse word “damn”, the evaluative adverbials “unfortunately” and “fortunately”, or the interjections “boo” and “yay”); this comparison can inform questions such as whether face emoji scopally act with operators in the accompanying text, e.g. giving rise to projective inferences.

April, 8th, 2021, Marloes Oomen

Neg-raising constructions in French Sign Language

In Neg-raising constructions, a negation which is realized in the main clause of a complex construction logically seems to belong to the dependent clause. For instance, I don’t believe that p is usually interpreted as conveying I believe that ¬p. Having garnered much theoretical interest for its potential to be accounted for from both syntactic and semantic/pragmatic perspectives (Horn 1978; Collins & Postal 2014; Gajewski 2007), the phenomenon of Neg-raising has been well-studied in spoken languages. In contrast, it has hardly been investigated in signed languages at all. In this talk, we report on our study of Neg-raising constructions in French Sign Language (LSF). Through the application of a variety of tests, we demonstrate that Neg-raising readings exist in LSF. We show that Neg-raising constructions have properties similar to those described for spoken languages, most notably regarding their interaction with the sign until, which we show is a strong NPI equivalent to e.g. English punctual until (cf. Lakoff 1969).

April, 1st, 2021, Tobias Gerstenberg

The counterfactual simulation model of causal judgment: A novel approach to understanding causal verbs

How do we make causal judgments? And how do we choose from a variety of causal expressions the one that best describes what happened? I will present a computational framework for modeling causal explanations in terms of counterfactual contrasts defined over intuitive domain theories. Focusing on people's intuitive understanding of physics, I will present several lines of experiments testing this framework. The counterfactual simulation model (CSM) predicts people's causal judgments about a variety of physical scenes, including dynamic collision events, complex situations that involve multiple causes, omissions as causes, and causal responsibility for a system's stability. It also captures the cognitive processes underlying these judgments as revealed by spontaneous eye movements. I will discuss how the CSM may help us better understand the mapping between causal events in the world and the words we use to describe them.

March, 25th, 2021, Wesley Webb (animal communication)

New horizons for female birdsong: evolution, culture and analysis tools

Female birdsong is widespread but poorly understood. In this talk I will share my doctoral research, which attempted to address the female song research gap in three ways.

First, I examined the evolution of female song across the songbird group, and how it relates to female plumage colour. I found that female song and female colourfulness have evolved together. In species where females sing, females tend to be more colourful, and these two traits probably have similar functions.

Next, I worked with a computer scientist to create software for classifying and analysing birdsong and other animal sounds. This free tool (Koe – koe.io.ac.nz) allows you to quickly sort song units into types so that you can study repertoire and sequence structure.

Koe was crucial for my field study on song culture in New Zealand bellbirds (Anthornis melanura). My team recorded wild bellbird song at six beautiful sites in the Hauraki Gulf, New Zealand. Comparing vocal diversity between sexes and sites, I found that male and female population sectors had equal vocal diversity, but little male–female overlap in repertoire. The degree of repertoire sharing between sites was not significantly different for males and females, and was generally very low, suggesting that bellbirds learn local songs after immigrating.

The thesis reveals that female song can be as elaborate as male song, and that male and female song cultures can be largely distinct. Further research is needed to explain the mechanisms underlying culture differences between sexes.

March, 25th, 2021, Jakub Dotlacil

Can a rational theory of memory model parsing?

This talk explores how the rational theory of memory summarized in Anderson (1991) can inform models of human parsing. It is shown that transition-based parsing is particularly suitable to be combined with Anderson's theory of memory systems. The combination of the rational theory of memory with the transition-based parsers results in a model of sentence processing that is data-driven, can be embedded in a cognitive architecture and informed by other findings related to memory and recall from memory. In the talk, I present the gist of the idea, the general framework and show how this approach allows us to model on-line behavioral measures (reading time data, regressions in eye-tracking studies).

March, 18th, 2021, Janek Guerrini

Similarity statements as a window into lexical meaning

What do we mean when we say of a statue of a lion that it looks like a lion? The statue resembles a lion in some specific respects: for one, it has the shape of a lion. But is this a derivative fact, or is "seem" actually sensitive to particular features of the concept of a lion?

The answer to this question has important consequences for lexical semantics. In psychology, some major proposals hold that conceptual representations are organized around features of some sort. In semantics, the question of the nature of lexical items is still open. Traditionally, lexical items like "lion" have been analyzed as extensions, viz. the set of all lions.

I make two points on the semantics of “seem": (i) because “seem" is grammatically sensitive to a featural structure, lexical items cannot be mere extensions. (ii) Not just any richer ontology will do: the behaviour of lexical items under “seem" needs to be captured in terms of attribute-value pairs.

More specifically, (i) I examine potential modal analyses of “seem" combined with an extensional lexical semantics and I show that they do not correctly predict the failures of distributivity displayed by "seem": if something seems like a A&B, it does not necessarily seem like a A and seem like a B. (ii) I then examine various ways to regiment the internal structure of lexical items. Viewing them as mere collections of features is not enough to capture the behavior of "seem". One way to add structure to features is to represent lexical items as collections of attribute-value pairs (e.g. color-blue, texture-soft). I argue that this ontology does best at capturing the fact that “seem" is sensitive  to attributes rather than unstructured features: we can say that x looks like a lion "with respect to color", where “color” cannot be viewed as a full property.

February, 25th, 2021, Julian Grove

Presupposition projection as a scope phenomenon

Heim's satisfaction theory of presupposition projection has paved the way for a successful research tradition within dynamic semantics that has been able to give compositional analyses of a variety of presupposition projection behaviors. Since Geurts 1996, however, the satisfaction account has been known to lead to incorrect predictions in cases where a presupposition ends up filtered which should not have been. I show that the satisfaction account of presupposition projection can meet Geurts's challenge by revealing that his observations are valid only under certain basic assumptions about semantic composition. To illustrate, I slightly sophisticate the compositional repertoire by allowing presupposition triggers to take exceptional scope, in a manner reminiscent of mechanisms proposed for indefinites. Doing so leads to a setting in which the proviso problem is no longer present. If time permits, I'll also demonstrate how the proposed system can be extended to handle binding and presupposition accommodation.

February 18th, 2021, Kyle Blumberg & Simon Goldstein

A Grammatical Theory of Redundancy This paper develops a grammatical theory of redundancy. The operator Red(p) says that p is true and any sentence simpler than p is distinct from p in Red(p)'s local context. Red can occur under the scope of quantifiers and other operators. This operator explains a variety of empirical data, including diversity conditions, ignorance inferences, and free choice. In particular, forms like 'Alex fears Red(it is raining)' explain why attitude verbs require the agent's beliefs to be diverse regarding their complement. Forms like 'Alex hopes Red(Billie or Carol is at the party)' explain why complex attitude ascriptions imply ignorance of each constituent. Forms like 'Alex is permitted to Red(eat cake or ice cream)' explain why disjunctive possibility and necessity claims imply that each disjunct is possible. The grammatical theory of redundancy compares interestingly with the grammatical theory of scalar implicature.

February, 11th, 2021, Mora Maldonado

Ain't got no: An artificial language learning approach to negative dependencies and concord

Languages vary with respect to whether sentences with two negative elements give rise to double negation (DN) or negative concord (NC) meanings. For example, in Dutch, the combination of the negative indefinite "niemand" and of the negative marker "niet" results in a positive meaning, as one can see in (1):

(1) Niemand rent niet.

N-body run NEG.

“Nobody doesn’t run” → “Everybody runs”

This is not the case in all languages. For example, despite involving two negative elements, a Serbian sentence like (2) only contains one semantic negation. In other words, the two negative elements in (2) do not each contribute a semantic negation, but rather convey a negative meaning together, yielding what is known as a negative concord interpretation.

2) Niko ne trci

N-word NEG run

"Nobody runs"

We explore an influential hypothesis about what governs this variation: namely, that whether a language exhibits DN or NC is partly determined by the phonological and grammatical nature of its negative marker (Jespersen, 1917; Zeijlstra 2004).

We ran two artificial language experiments to investigate whether learners are sensitive to the correlation between negative marker and interpretation. Are English-speakers more likely to treat a language as negative concord if the negative marker is an affix than if it’s an adverb? What about treating it as double negation? Our findings fail to provide learning evidence for the connection between type of negative marker and type of language. However, our results suggest that participants find it easier to learn negative concord languages independently of whether the negative marker is an adverb or an affix. Moreover, learners seem to be sensitive to the input variability, such that they need very little evidence to adjudicate between type of language.

February, 4th, 2021, Milica Denić

Complexity/informativeness trade-off in the domain of indefinite pronouns

The vocabulary of human languages has been argued to support efficient communication by optimizing the trade-off between complexity and informativeness (Kemp & Regier, 2012). The argument has been based on cross-linguistic analyses of vocabulary in semantic domains of content words such as kinship, color, and number terms. The present work extends this analysis to a category of function words: indefinite pronouns (e.g. someone, anyone, no-one, cf. Haspelmath, 2001). We establish the meaning space and feature-based representations for indefinite pronouns, and show that indefinite pronoun systems across languages optimize the complexity/informativeness trade-off. This demonstrates that pressures for efficient communication shape both content and function word categories, thus tying in with the conclusions of recent work on quantifiers (Steinert-Threlkeld, 2019). Furthermore, we argue that the trade-off may explain some of the universal properties of indefinite pronouns, thus reducing the explanatory load for linguistic theories.

January, 21st, 2021, Nur Lan

Small Neural Networks for Better Generalizations

Recurrent neural networks (RNNs) are the go-to technology for a number of tasks, including linguistic tasks. They nonetheless face two well-known challenges related: (a) the difficulty of such networks to generalize appropriately as opposed to memorizing, especially from very short input sequences (generalization); and (b) the difficulty for us to understand the knowledge that the network has attained (transparency). We propose to change the way such networks are trained, incentivizing them not only to be accurate but also to be small. This is done in a principled way, using the so-called "Minimum Description Length" principle (MDL) as an objective function, a measure that is meant to balance size and accuracy, in a way Occam's razor would have it. We find that MDL leads the networks to reach adequate levels of generalization from very small corpora, improving over classical training regimes (backpropagation). We will go through linguistic-like artificial tasks where RNNs typically fail to match human generalization from surface forms, and show our current results in the domains of formal languages and semantics. With the MDL approach, the networks are able to learn underlying grammars in a transparent way.

December, 10th, 2020, Dilip Ninan

The Projection Problem for Predicates of Taste

Utterances of simple sentences containing taste predicates (e.g. "delicious", "fun'', "frightening'') typically imply that the speaker has had a particular sort of first-hand experience with the object of predication. For example, an utterance of "The carrot cake is delicious'' would typically imply that the speaker had actually tasted the cake in question, and is not, for example, merely basing her judgment on the testimony of others.

One puzzling aspect of this inference is its projection profile: although the acquaintance inference projects over negation, it fails to project out of other standard presupposition holes such as epistemic modals, the antecedents of conditionals, and questions. This curious behavior has given rise to two different theories. According to the first, the acquaintance inference is not a presuppositional inference, but an implicature, one generated by the Maxim of Quality together with a certain principle concerning the epistemology of taste (Ninan 2014). According to the second, the acquaintance inference is a presuppositional inference, but one that can be systematically obviated by certain kinds of operators (Anand & Korotkova 2018; Willer & Kennedy Forthcoming).

I will begin by discussing some problems for the implicature theory that arise in connection with disjunction and quantification. I'll then argue that a supervaluationist version of the `presupposition-plus-obviation' theory avoids these problems, and also handles the other data motivating the implicature approach. I'll then suggest that the supervaluationist theory is further supported when we examine its predictions concerning the interaction between the acquaintance requirement and various generalized quantifiers. Finally, I will turn to the question of which operators obviate and why.

December, 3rd, 2020, Mariapaola D'Imperio

The impact of individual pragmatic skills on the prosodic-meaning mapping

The pragmatic valence of intonation contours has been under investigation for various decades, though only recently studies have unveiled that both group- and speaker-specific strategies are employed at the level of meaning decoding in prosodically variable stimuli (e.g., Cangemi et al. 2015 for focus in German; Portes & German 2019 for yes/no questions in French; Warren 2016 for uptalk in New Zealand English). Variability in encoding intonational meaning has mainly focused on pitch contours in production. Co-speech gestures have also been recently found to variably mark information structure and appear to precede prosodic encoding in children (Estève-Gibert et al., to appear). However, research on the impact of listener’s cognitive differences on decoding intonation-meaning is still in its infancy.

In this talk I will offer evidence showing that the intonation-meaning mapping is not homogeneous within a language-community. Specifically, I will show the results of two recent studies on the role of pragmatic skills on recovering intonational meaning. I will first briefly show the results of an Eyetracking study on French intonation processing showing the impact of individual empathy scores on recovering contrastive meaning (Esteve-Gibert et al. 2020). I will then present a perception experiment in which the role of pitch range in the nuclear configuration region is tested as a predictor of type (negative vs positive) and degree of epistemic bias in Salerno Italian (Orrico & D’Imperio 2020). Here, two independent sources of individual variability were uncovered: Listeners’ prolonged exposure to either a different language or variety and their Empathy Quotient (EQ), which interacted in epistemic bias identification. I will then discuss the results in terms of viable theoretical models of intonation-meaning mapping.

November, 26th, 2020, Bogdan Dicher

The metainferential conception of logic: an opinionated survey

The (improperly called) metainferential conception of logic is the view that logical consequence relations are determined by the inferences that hold between a logic’s inferences, rather than by the inferences that hold between its formulae. In this talk I will review the motivation for advancing this view of logic, some of key ways in which it differs from the standard conception(s) of logic and some of the philosophical challenges that lie ahead for its proponents.

November, 19th, 2020, Keny Chatain & Amir Anvari

A theory of exhaustification based on Logical Integrity and Competence

Logical Integrity (LI) is a principle of language use that has recently been motivated (Anvari 2018) on the basis of a broad class of use-conditions associated with sentences which have logically non-weaker alternatives. Originally, this principle was formulated with reference to the common ground. In this talk, we begin with the observation that if the principle is formulated with reference to the speaker’s beliefs instead, it turns into a weak inferential engine: it predicts inferences that are typically weak versions of attested, and much discussed, inferences (such as classic scalar implicatures, distributive inferences, free choice, ignorance and so on). On the basis of this observation, we propose a stregnthening mechanism, COMP, which mirrors the competence assumption of neo-Griceans (particularly Spector 2007). We will argue that this combination of LI and COMP improves on extant theories of exhaustification (Fox 2007, and subsequent developments) in providing a uniform account of general distributive inferences associated with disjunction in various environments. Furthermore, our proposal in and of itself does not necessitate an implementation in terms of covert operators; it is compatible with the alternative view that certain patterns of reasoning (encoded by LI and COMP) are performed in local contexts (Schlenker 2009). We will discus some obstacles to the the radical, operator-free version of our theory, including Hurford's constraint. We will sketch a possible solution for Hurford's constraint based on the idea that the (local) accommodation operator turns the inferences generated by LI+COMP into the assertive content.

November, 12th, 2020, Deanna Gagne

The ball rolled to the right: When children accommodate parents in the absence of language peers

This presentation will discuss work relating to influences on the regularization of spatial reference in an emerging sign language. Oftentimes, changes within a linguistic system are assumed to be confined to the language input and the language output. This relationship is also typically conceptualized as vertical; the input is rich and flows from older individuals with mastery of the language to young learners who are mastering the language. Here, I explore a scenario in which the vertical relationship may not represent mastery and in which the horizontal, or peer relationship, may contribute more strongly to the emergence of linguistic convention. Further, I present data that demonstrates that language conventionalization may be influenced by factors beyond statistical patterning in the input; namely the linguistic-cognitive abilities of interlocutors in the system.

November, 5th, 2020, Lena Pasalskaya & Jeremy Kuhn

Negative concord in Russian Sign Language

In this talk, we provide the first systematic description of negative concord in Russian Sign Language (RSL). Although non-manual signs have been argued to participate in negative concord in sign languages, negative concord involving manual signs has been shown to be much rarer. The RSL pattern thus fills this typological gap, providing one of the first clear cases of negative concord of manual signs in sign language. We show that RSL fits into the typology of negative concord known from spoken language: RSL is a language with non-strict negative concord, in which specific structural positions allow NC items to appear without a licensor. On the other hand, RSL diverges from other non-strict concord languages with respect to important macro-syntactic properties: in RSL, like in a number of other sign languages, negative words may appear in a right-branching structure. These syntactic properties provide new evidence that structural hierarchy, not linear order, is responsible for explaining the presence or absence of a licensor in non-strict concord languages.

October, 29th, 2020, WooJin Chung

Modality, expected utility, and Bayesian confirmation

We introduce an expected value theory of linguistic modality that makes reference to expected utility and likelihood-based confirmation measure for deontics and epistemics, respectively. We derive the modal semantics in an entirely transparent manner, as it is based on the compositional semantics of Korean modal expressions that are morphosyntactically decomposed into a conditional and an evaluative predicate. We show that our theory is not only theoretically advantageous but also has interesting empirical consequences: it suggests that people's conception of modality facilitates rational decision making, but the very same mechanism is the source of irrationality in assessing comparative likelihoods.

October, 22nd, 2020, Jeremy Kuhn

Disjunctive discourse referents in sign language (work in progress)

Disjoined noun phrases may introduce disjunctive discourse referents. In (1), the pronoun refers to whichever individual Mary sees. This is perhaps expected, as disjunction can be seen as indefinite quantification over a domain of two individuals. Disjunction provides more flexibility in some respects than indefinites, though, as the two disjuncts may bear different morphological features, and a disjunctive discourse referent may have a split antecedent, as in (2).

(1) Whenever Mary sees a donkey or a horse, she waves to it. (Elbourne 2005)

(2) Whenever Mary catches a fish or John traps a rabbit, Bill cooks it. (Stone 1992)

Sign language has been shown to bear on arguments pertaining to discourse anaphora. Notably, discourse referents may be established at locations in the signing space, closely paralleling the use of variables in dynamic semantics (Schlenker 2011). This relation has been shown to be mediated by a "featural" layer, behaving similarly to phi-features like gender (Kuhn 2016). In this talk, we compare several theories of disjunctive anaphora and of space in sign language with new data from French Sign Language.

October, 15th, 2020, Nadine Bade

Alternatives and attention in language and reasoning: the case of modals

In this paper, we employ a novel experimental paradigm using insights from the psychology of reasoning to investigate the question whether certain modals generate and draw attention to alternatives. We present two experimental studies on English modals "might" and "allowed to" suggesting that whereas epistemic "might" is conventionally associated with alternatives, deontic "allowed to" is not. Based on our findings, we, first, discuss theoretical consequences for existing accounts of weak modality and, second, argue that using reasoning tasks is a powerful diagnostic tool for exploring the issue of which expressions involve alternatives.

October, 8th, 2020, Paul Egré & Benjamin Spector

On the optimality of Vagueness: "Around", "Between" and the Gricean Maxims

(joint work with Adèle Mortier and Steven Verheyen)

Why is our language vague? We argue that in contexts in which a cooperative speaker is not perfectly informed about the world, the use of vague expressions can offer an optimal tradeoff between truthfulness (Gricean Quality) and informativeness (Gricean Quantity). Focusing on expressions of approximation such as ``around'', which are semantically vague, we show that they allow the speaker to convey indirect probabilistic information, in a way that gives the listener a more accurate representation of the information available to the speaker than any more precise expression would (intervals of the form ``between''). We give a probabilistic treatment of the interpretation of ``around'', and offer a model for the interpretation and use of ``around''-statements within the Rational Speech Act (RSA) framework. In our account the shape of the speaker's distribution matters in ways not predicted by the Lexical Uncertainty model standardly used in the RSA framework for vague predicates. We draw further lessons concerning the semantic flexibility of vague expressions.

July, 23rd, 2020, Markus Neuwirth

Mapping Conditions and Hierarchicality in Music Semantics

Owing to recent work by Swain, Cook, Schlenker, and others, the „disaster area“ (Longuet-Higgins) of music semantics has turned into a promising and growing field of research. My talk contributes to this trend, aiming to reconsider the conditions for (plausible) music-world mappings, the interaction of musical properties, and the case of deeper-level iconicity. Each of these aspects will be illustrated by analytical close-readings of musical examples. By way of conclusion, I will briefly sketch an agenda that may potentially guide future research on music semantics.

July, 2nd, 2020

Anastasia Tsilia

Sequence of tense in Modern Greek

Are temporal features always semantically interpreted? Imagine that John took the following decision a week ago: ‘In 10 days, I will say to my mother that we are having our last meal together’. As Abusch (1988) points out this would be reported with a past-under-past in English, as in John decided a week ago that in 10 days he would say to his mother that they were having their last meal together. We get a simultaneous reading, according to which the most embedded past is neither prior to ‘our now’ nor to ‘John’s local now’. Therefore it cannot be interpreted as a past. This would be explained by the presence of a Sequence of Tense (SOT) rule in English. However, if the very same sentence was uttered in Hebrew the embedded past would be interpreted and we would end up understanding that John’s utterance will be ‘We were having our last meal’ (Sharvit 2003). That is because Hebrew has no such SOT rule. How would the targeted reading be expressed in a non-SOT language? The answer depends on whether the present tense is shiftable, i.e. may refer to the attitude holder’s local now, or not. In Hebrew this is indeed the case, as we may say 2000 years ago, Yosef found out that Miriam loves him (Ogihara & Sharvit 2012). Modern Greek seems to behave like Hebrew in this respect. We are going to further establish the generalisation found in Schlenker (1999) and Sharvit (2003; 2017), namely that Modern Greek has optional SOT, allowing for both a present- and a past-under-past to get a simultaneous reading. In this talk I will present my work with Philippe Schlenker and Amir Anvari this semester, establishing that there is a SOT rule and that the present tense is shiftable under past tense attitude verbs. I will sketch a possible theoretical explanation, based on a general preference for de se Logical Forms. However, the same does not hold for relative clauses, where the present tense does not shift. Nevertheless, we noticed with Amir Anvari that there are some predicates that allow for shiftable tense there too (such as ‘see’, ‘look for’, ‘meet’, ‘talk to’). I am going to present these data and, time permitting, talk about another puzzle: future-under-future yielding simultaneous readings in Modern Greek, even though no SOT rule is supposed to hold in such a case.

Amir Anvari

On in two days

I report on ongoing work with Janek Guerrini and David Blunier pertaining to the semantics of temporal adverbs "two days ago" and "in two days" in Farsi, Romagnol, Italian, French, German. Our goal is to understand the similarities and the differences between the interpretation of these two adverbs, on the one hand, and indexical adverbs like "the day after tomorrow" / "the day before yesterday" and anaphoric adverbs like "two days later" / "two days before", on the other.

June, 25th, 2020, Amir Anvari

The variable-based account of indexical shift redux

The observation that in some languages indexicals can shift, i.e., can refer to objects other than those that make up the actual speech event, was something of a bombshell thrown by Philippe Schlenker in his 1999 dissertation, with quite far reaching consequences both for philosophy of language and linguistic analysis proper. Schlenker took the data to be pointing to a view of language on which attitudinal predicates quantify over contexts and indexicals involve context variables that, when bound by attitudes, shift. However, in subsequent work, particularly by Anand 2006 and Deal 2017, a host of cross-linguistic generalizations have been uncovered that suggest a tight typology of shifty indexicals, one that does not seem to justify the massive expressive power of Schlenker's extensional proposal. Anand and Deal, therefore, have proposed an intensional analysis of the facts in which context-shifting operators, chained by an array of syntactic stipulations, may overwrite the context parameter and shift indexicals in the process, but in a manner restricted enough to handle the typology. In this talk, I will sketch the core typological generalizations and the operator-based analysis, basing the discussion on Deal 2017. Then, building on ongoing work with Janek Guierrini and David Blunier, I will discuss some novel generalizations pertaining to the Farsi temporal adverb "do ruz dige" (lit. two days other, similar to "in two days" in English) which pose a significant challenge to the operator-based approach. On this basis, I propose a return to the variable-based analysis. In a very loose sense, my proposal will be seen to be a translation of the intensional analysis into a multi-sorted extensional type theory. A key role will be played by what I will refer to as Zimmermann's technique (p.c. to Philippe Schlenker, reported in Schlenker 2009 pg. 28). Unlike its predecessors (Schlenker 1999, 2003, von Stechow 2002), this proposal matches the predictions of the operator-based approach, and therefore is not susceptible to the over-generation criticism, while at the same leaving enough room for a simple analysis of "do ruz dige".

June, 18th, 2020, Salvador Mascarenhas

Lawyers-and-engineers revisited: base-rate neglect and the inverse fallacy in reasoning with alternatives

Irrationality observed in tasks such as the lawyers-and-engineers problem (Kahneman and Tversky 1973, Psychology of Prediction) has been used to argue that humans systematically disregard the prior probabilities of events in reasoning and decision making. In this and related reasoning tasks, subjects learn about a simple probability distribution regarding two professions (70% lawyers and 30% engineers). Then they are presented with individuating information about a randomly chosen individual from that population (Bill dislikes public speaking, he enjoys gardening and logic puzzles). When asked to estimate the probability that the individual belongs to one of the two professions, subjects appear to ignore prior probabilities and respond entirely as a function of the individuating information. This effect has been interpreted in terms of representativeness, a heuristic argued to substitute actual reasoning about probabilities.

I show that confirmation-theoretic reasoning, which yields insightful accounts of the conjunction fallacy (Tentori, Crupi, Russo, 2013, On the determinants of the conjunction fallacy. JEP) and of some

deductive reasoning fallacies, extends to the lawyers-and-engineers problem. When faced with a question that sets up multiple possible answers (hypotheses), reasoners pick the one that best explains available information (evidence).

Kahneman & Tversky's original experiment design was well equipped to reject a purely rational theory involving solely the posterior probabilities of the possible answers, but it could not distinguish

non-probabilistic similarity judgments (representativeness) from sophisticated (probabilistic) confirmation.

In joint work with Janek Guerrini, I collected the components required to compare confirmation-theoretic models in a lawyers-and-engineers setting: priors (hypotheses and evidence), posteriors, and likelihoods. We used pairs of real-world professions (hypotheses) and related one-sentence descriptions (evidence) in a norming study, asking participants for frequency judgments. An independent group solved lawyers-and-engineers problems with the normed materials. We found more rational behavior than previous studies, and little evidence of prior-probability neglect. Models leveraging likelihoods (i.e. P(description | profession)) in addition to posteriors outperformed posteriors-alone models. In fact, the best models for our data are based on the log ratio of the likelihoods for the two professions, showing that humans follow a comparative strategy when

solving these problems. This finding confirms the predictions of the erotetic theory of reasoning (Koralus and Mascarenhas, 2013, Phil Perspectives).

In another study (joint work with Mathias Sablé-Meyer), we assessed whether familiar stereotypes and complex descriptions are essential features of the phenomenon, as is predicted by the representativeness account. We distilled the task to its essence and eliminated linguistic elements as much as possible, conveying probabilistic distributions visually by displaying objects of two shapes

(hypotheses) in one of two colors (evidence). In each trial, participants were told that an element had been picked randomly and were given its color, then had to guess its shape. We found the same

results as in the first experiment, suggesting that this reasoning process is independent of familiarity and does not require individuating information. At this point, we cannot exclude the possibility that rich-enough stereotypes are formed on the fly.

June, 11th, 2020, Edgar Onea

Focus Exhaustivity is independent of Bayesian reasoning

One of the main recent advances in formal pragmatics is the Rational Speech Act Model, that has proven successful in the modeling of a number of pragmatic inferences based on a recursive Bayesian reasoning. Exhaustivity inferences have been particularly often studied in this framework with remarkable success. In this talk we present experimental evidence suggesting that focus exhaustivity does not easily lend itself for an analysis withing the RSA framework; in fact the RSA framework does not seem to be suitable to derive the exhaustive reading associated with narrow focus at all. Instead, the role of the RSA framework seems to be limited to optimizing between an exhaustive and non-exhaustive interpretation, whereby exhaustivity needs to be independently derived.

June, 4th, 2020, Jeremy Kuhn

Tense and aspect in LSF

While many sign languages do not have overt morphological tense marking, most have been shown to have rich aspectual systems, including both lexical items and iconic phonological modifications of a verb form (Rathmann 2005, i.a.). Some of these iconic strategies have been the center of theoretical debate, including 'end-marking' on verbs, which has been shown to be correlated with telicity (Wilbur 2003, Strickland et al. 2015). In this talk, I discuss a variety of strategies for marking tense and aspect in French Sign Language (LSF), with the following general findings. Primary tense is not morphologically marked in LSF. Secondary tense is marked with lexical items FINI (perfect) and VA-VA (prospective), as well as possibly some suppletive verb forms. Contra Davidson et al. (2019), end-marking does not mark perfective aspect, but a variety of repetition-based modifications may mark flavors of imperfective aspect, including progressive and iterative/habitual. I show that these aspectual modifications are both iconic and scope-taking, showing new cases of "scopable iconicity," as proposed in Kuhn & Aristodemo (2017). Regarding 'inner aspect' (aktionsart), I show that the telicity of a predicate is determined by properties of both the verb its arguments. One such property is iconic: focusing on degree achievements, I propose that the phonological form of verbs is iconically mapped to event progression, with end-marking associated with a scalar maximum.

May, 28th, 2020, Matthew Mandelkern

Good Guesses

(joint work with Kevin Dorst)

We explore the practice of guessing: how subjects make predictions amongst several optionswhen they can’t pick one with certainty. Guesses show intricate and surprising patterns that themost obvious theories do not predict. We give a model which does predict these features based onthe Jamesian idea that subjects aim to optimize a trade-off between accuracy and informativityin their answers. We then use this model to argue that guessing has a central role to play in ourcognitive and conversational lives. In particular, we argue that the accuracy-informativity tradeoffthat governs guessing offers (1) a theory of belief, (2) a generalization of the standard pragmaticsof assertion, and (3) an account of the conjunction fallacy—the puzzling psychological findingthat people sometimes rate conjunctions as more probable than their conjuncts.

May, 19th, 2020, Frédérique de Vignemont

Keeping the world at distance

Social distancing is not only the result of our complex modern societies at a time of disease. Rather it reveals an evolutionary old sensorimotor mechanism, known as peripersonal space, which can be conceived as a last resort margin of safety. The machinery relies on a basic principle. What is close to our body can be soon in contact with it, either because the close-by object moves or because we move. A further basic fact is that this contact may be welcome (when grasping an object), but not necessarily (when colliding into an obstacle). Hence, there is an immediate significance of our surrounding, which imposes clear rules for its perception. In brief, peripersonal perception must follow the old Scout’s motto: “Be prepared” (semper paratus). It must be prepared both at the sensory level (prepared to detect and process whatever may be soon on the skin) and at the motor level (prepared to react to it). Most of the time, despite these computational specificities, we seem to have no conscious awareness of peripersonal space as being “special” in any sense. Instead we are presented with a continuous visual field without a phenomenological boundary between what is close and what is far. However, when what is close to us is affectively loaded, like someone coughing next to us, we anticipate her contact and this anticipation makes us feel uncomfortable. I will argue that this feeling of discomfort result from tactile expectation.

“A number of porcupines huddled together for warmth on a cold day in winter; but, as they began to prick one another with their quills, they were obliged to disperse. However the cold drove them together again, when just the same thing happened. At last, after many turns of huddling and dispersing, they discovered that they would be best off by remaining at a little distance from one another. In the same way the need of society drives the human porcupines together, only to be mutually repelled by the many prickly and disagreeable qualities of their nature. The moderate distance which they at last discover to be the only tolerable condition of intercourse, is the code of politeness and fine manners; and those who transgress it are roughly told—in the English phrase—to keep their distance. By this arrangement the mutual need of warmth is only very moderately satisfied; but then people do not get pricked. A man who has some heat in himself prefers to remain outside, where he will neither prick other people nor get pricked himself.” (Schopenhauer, 1851)

May, 14th, 2020, Amir Anvari

Copy theory of movement and proper treatment of intensionality On the classical view, the interpretation function is equipped with a pair of parameters, the context and the index, relative to which expressions are mapped to their extensions; the so-called "indexical" expressions (e.g., the first person pronoun) are dependent on the context parameter while almost all other expressions (e.g., predicates) are dependent on the index parameter. On top of this, the classical view is coupled with the assumptions that modal operators "shift" the index parameter, and that no operator shifts the context parameter (Kaplan 1977). Two predictions immediately follow, that index-dependent expressions in the scope of modal operators are obligatorily interpreted De Dicto while context-dependent expressions are always interpreted De Re. Both predictions are incorrect (Schlenker 1999, Percus 2000), and there has been a steady stream of attempts in the literature to address these problems without veering too far away from the main tenets of the classical view (i.a., Anand 2006, Deal 2017, Keshet 2008, Demirok 2019). In this talk, I will argue that copy theory of movement (Chomsky 1995; Takahashi 2010 for recent overview) poses a difficult challenge to these attempts. If movement leaves behind, not a trace, but a full-fledged copy of the moved expression, then any time movement crosses a modal operator, the system generates unattested "double-access" readings: the higher copy, begin outside of the scope of the modal, will be interpreted De Re while the lower copy, being in the scope of the modal, will be interpreted De Dicto. Against this background, I will point out that the extensional alternative to the classical view, i.e., the family of frameworks in which contexts or indices are referred to in the object language, while suffering from a massive over-generation problem in general (Anand 2006, Deal 2017, Percus 2018), is nevertheless straightforwardly compatible with copy theory with the aid of a handful of rather plausible assumptions.

May, 7th, 2020, Diego Feinmann

The Character of Imprecision

In this talk, I will be concerned with the phenomenonof imprecision, also referred to as loose talk or pragmatic slack (Lasersohn1999). To begin with, I will discuss what I take to be the main features of imprecision.Second, following Križ (2015), I will claim that there is a link between homogeneity and imprecision: specifically, that homogeneity is a pre-requisitefor imprecise interpretation. In the final part of the talk, I will conceptually and empirically motivate the notion of ‘invisible homogeneity’,discuss the significance of this notion in the context Križ’s (2015) pragmatic framework, as well as put forward a homogenous semantics for absolute adjectives.

May, 5th, 2020, Franck Ramus [LINGUAE+ seminar]

Genetics & Education

Education level is a major determinant of employment, income, health and life expectancy. In turn, the most important determinant of educational achievement is thought to be parental socioeconomic status. Is it? To what extent does the observed correlation reflect causation? We will review old and new evidence for the impact of genetic factors on educational achievement, and discuss their potential implications.

Link video

April, 30th, 2020, Kajsa Djarv

Decomposing factivity: presupposition, evidence, and assertion

In this talk I examine the claim that factive verbs, like know, discover, and resent, presuppose their embedded proposition (p). I focus on three dimensions of the factive presupposition: projection, entailment, and assertability; investigating each of these dimensions in the context of the split between cognitive and emotive factives. From this data, I argue for three main conclusions; first, that neither cognitive nor emotive factives presuppose p; second, that we need an analysis of factive verbs that separates the (generally projective) p-inference from the status of p with respect to the Question Under Discussion; and third, that an account which provides a unified source of the p-inference for cognitive and emotive factives is desirable. Finally, I evaluate different alternatives for such an analysis, ending with some conclusions about what cognitive and emotive factives do and do not presuppose.

April, 23rd, 2020, Philippe Schlenker (joint work with Jonathan Lamberton)

Iconic Plurals

In several sign languages, in some homesigns, and in some speech-replacing gestures, plurals can be marked by repetitions, with a distinction between punctuated and unpunctuated repetitions, which respectively come with precise and vague quantitative conditions  (Pfau and Steinbach 2006, Coppola et al. 2013, Abner et al. 2015).  Schlenker and Lamberton 2019 show that ASL plurals can have an irreducibly iconic component, e.g. an unpunctuated repetition of TROPHY arranged as a triangle may serve to refer to trophies with a triangular arrangement; but they took as primitive that repetitions serve to express plurals. 

Here we provide an iconic semantics for the repetitions themselves, using the formal semantics for pictorial representations developed in Greenberg 2013. Simplifying somewhat: punctuated repetitions correspond to precise, 'exactly' quantitative conditions because they are easy to count; unpunctuated repetitions correspond to imprecise, 'at least' quantitative conditions because they are  presented as hard to count and are blurry (vague) representations. Our analysis has three main modules: (i) a formal semantics for pictorial representations, (ii) a vagueness component (a simplified version of ‘tolerant’ approaches), and (iii) a mechanism of pragmatic exploitation (within a simplified RSA model), which is essential to obtain ‘at least’ readings for unpunctuated repetitions.

April, 21st, 2020, Jean-François Bonnefon [LINGUAE+ seminar]

Universals and variations in moral decisions made in 42 countries by 70,000 participants

When do people find it acceptable to sacrifice one life to save many? Cross-cultural studies suggested a complex pattern of universals and variations in the way people approach this question, but data were often based on small samples from a small number of countries outside of the Western world. Here we analyze responses to three sacrificial dilemmas by 70,000 participants in 10 languages and 42 countries. In every country, the three dilemmas displayed the same qualitative ordering of sacrifice acceptability, suggesting that this ordering is best explained by basic cognitive processes rather than cultural norms. The quantitative acceptability of each sacrifice, however, showed substantial country-level variations. We show that low relational mobility (where people are more cautious about not alienating their current social partners) is strongly associated with the rejection of sacrifices for the greater good (especially for Eastern countries), which may be explained by the signaling value of this rejection.

April, 16th, 2020, Danny Fox

Towards a Uniform Semantics for Question Embedding

While partitions provide the right kind of object for the characterization of the pragmatics of questions, they do not provide the right denotation to account for the semantics of question

embedding. This conceptual challenge is accompanied by the empirical challenge of accounting for the different types of schemas that have been identified for question embedding, in particular, “strong exhaustivity”, “weak exhaustivity” and “intermediate exhaustivity”. In this talk I will suggest that a trivalent definition of the exhaustivity operator (argued for by Bassi, Del Pinal and Sauerland) provides a way to deal with both of these challenges. Appling the operator to each member of the question denotation, provides the required partition, as I’ve tried to argue in recent work. Here I will argue that once the trivalent definition is adopted, independent observations about presupposition projection yield the different types of exhaustivity that are identified in different environments.

April, 9th, 2020, Emile Enguehard

Presupposition projection in (coordinations of) polar questions

Examples such as (1) and (2) show that in certain cases, presuppositions triggered by an element inside a question nucleus may fail to project. In fact, in what looks like coordinations of polar questions, presupposition projection patterns are exactly parallel to what is observed when the corresponding assertions are coordinated.

    (1) Is Ann married and is her husband abroad?

    (2) Is Ann single or is her husband abroad?

I will show that these facts do not fall out straightforwardly from existing theories of polar questions, (apparent) coordinations of questions, and presupposition projection. In fact, if we wish to capitalize on existing views about presupposition projection, we need a theory of questions such that (a) apparent coordinations of questions actually are coordinations of questions and (b) the semantic denotation of polar questions is asymmetric with respect to the “yes” and “no” answers. I will develop a trivalent theory of questions along these principles, such that the felicity of (1) and (2), and the infelicity of (3) due to redundancy are correctly predicted.

    (3) Is Ann in Paris and is she in London?

March, 26th, 2020, Leo Migotti

Research directions and perspectives in music semantics

This talk will be an opportunity to discuss my thesis directions on music semantics and aims at getting feedback. After describing the theoreticalframework of music semantics, I will focus on new possible semantic correlations involving musical parameters (such as pitch range,tempo, or loudness) that convey meaning in music, using both simplified and real music, as well as preliminary experimental results. For instance, it seems that a combination of conditions on both rhythmic and harmonic patterns can trigger fine-grained inferences about how a denoted object is moving.  Tests of these correlations could involve (i) a detailed analysis of real music, or (ii) large-scale analyses using computational tools applied to musical excerpts from different musical traditions. Besides systematic testing of these correlations, the semantics of music can also be revealed by analyzing its behavior when embedded in sentences.Recent experimental results suggest that simple iconic stimuli such as scales can easily trigger presuppositions as they are interpreted as changes of state.The case of cosuppositions (i.e. conditionalized presuppositions typicaly triggered by co-speech gestures), however, requires more complex stimuli, such as audiovisual animations or co-film music. In both cases, one can ask whether the same phenomenon arises without embedding.

March, 19th, 2020, Vincenzo Crupi

The logic, semantics, and psychology of evidential conditionals

Once upon a time, some thought that indicative conditionals could be effectively analyzed as material conditionals. Nowadays, an alternative theoretical construct prevails and receives wide acceptance, namely, the conditional probability of the consequent given the antecedent. Partly following earlier critical remarks made by others (most notably, Igor Douven), I advocate a revision of this consensus and suggest that incremental probabilistic support (rather than conditional probability alone) is key to the understanding of indicative conditionals and their role in human reasoning. There have been motivated concerns that a theory of such evidential conditionals (unlike their more traditional suppositional counterparts) can not generate a sufficiently interesting logical system. I will present results largely dispelling these worries.

Happily, and perhaps surprisingly, appropriate technical variations of Ernst Adams’s classical approach allow for the construction of a new logic of evidential conditionals which is nicely superclassical, fairly strong, restrictedly connexive, and potentially of consequence for empirical investigation. Moreover, much like in the case of Adams’s suppositional conditional, a parallel possible-world account is also available, so that the same logical profile is preserved and made consistent with a truth-conditional semantics. [Joint work with Andrea Iacona.]

March, 5th, 2020, Damian Szmuc

Immune Logics

In the past few years, the family of many-valued logics called infectious logics received an increasing amount of attention. These systems count with a truth-value that is assigned to a complex formula whenever it is assigned to some of its components---thus, behaving in an infectious way. Rather informally, we could say that these values behave in a "value-in-value-out" fashion. From a mathematical point of view, infectious values of this sort can be thought of as all-purpose zero elements. The aim of this talk is to discuss a family of many-valued logics that can perhaps be considered as duals to the infectious systems---whence, they will be called immune logics. In this vein, these logics count with a truth-value that is never assigned to a complex formula whenever it is assigned to some of its components, except in certain cases. Once again rather informally, we could say that in some of these cases these values behave in a "value-in-different-value-out" manner. Therefore, immune values of this sort can be thought of as all-purpose identity elements. As regards immune logics, our goal is to describe and analyze various three-valued systems. For this purpose, we explore immune logics where validity is defined by letting the immune value be designated, systems where it is undesignated, and systems where mixed notions of validity are adopted. In doing so, we highlight the links to various logics that have already appeared in the literature and some which were not discussed until now.

February, 27th, 2020, Gregory Bochner

A Puzzle about Demonstrative Reference in Stalnaker’s Two-Dimensional Pragmatics

My aims in this talk will be (i) to cast light on two features (which I will call "Presuppositionality" and "Detachability") of Stalnaker’s two-dimensionalist pragmatic account of assertion, (ii) to formulate two arguments, one from error and one from ignorance, relying on those two features to create a problem for this classic account, and (iii) to offer a diagnosis of the problem. Just as Kripke’s arguments from error and ignorance invalidated Frege’s one-dimensional Descriptivism, I will suggest, my related arguments here will invalidate Stalnaker’s two-dimensional Descriptivism. I will argue that the problem arises from the combination, in Stalnaker’s account, of Presuppositionality and Propositionalism, the doctrine that the contents of assertions and of pragmatic presuppositions are always propositions, which can be true or false only with respect to (entire) possible worlds.

February, 20th, 2020, Elena Pagliarini

Interaction between disjunction and negation from a cross linguistic and a developmental perspective

I will illustrate how children face the challenge of mapping sentences onto meaning by looking at the acquisition of disjunctionunder negation across different languages. Previous studies reached the conclusion that children start by preferring the subset of the truth-conditionsassigned to the negated disjunction according to classical logic, independentlyfrom the interpretation assigned by adult speakers of the local language(Crain, 2012; Geçkin et al., 2016; Goro & Akiba, 2004; Goro, 2007, a.o).This pattern of responses has been explained in terms of a principle oflanguage learnability called Semantic Subset Principle (Crain et al., 1994).

In my talk, I will outline a series of new studies from Italian,French, Hungarian, Dutch and Catalan. The results reveal a very complex picturewhich discloses some limits of previously suggested explanations. I will discuss the findings in light of linguistic differences ofthe negative concord system across languages and of possible developmentaldifferences among children.

February, 13th, 2020, Chierchia Gennaro

Homogeneity, scopelessness and discourse dynamics

There are three phenomena that appear to be intuitively related. They are homogeneity, which gets discussed predominantly in connection with definites, scopelessness, which concerns bare plurals, and the distribution of readings of donkey pronouns. In all of these cases the

reading one finds in negative environments is NOT what one would expect on the basis of the reading one gets in the corresponding positive environment. While these issues have been widely discussed separately, no unified theory of them exists at present, in so far as I know. My

goal is to develop such a theory, by viewing each of these phenomena as a case, ultimately, of free choice.

February, 6th, 2020, Yoad Winter

Presupposition Filtering: Trigger-Orientation vs. Proposition-Orientation in Trivalent Semantics

Theories of presuppositions struggle to account for contrasts like the following:

  (1) If Sue is in a good mood, she will bring her partner to the party

  (2) If Sue is in a relationship, she will bring her partner to the party

While (1) presupposes that Sue has a partner, (2) does not. The most popular approaches to presupposition projection - trivalent and satisfaction-based accounts alike - correctly expect the "filtering" in (2), but wrongly expect (1) to presuppose that Sue has a partner *if she is in a good mood*. This kind of undesirable prediction, known as the "proviso problem", is avoided by Karttunen's original filtering conditions (1973), though the semantic justification for such conditions has remained questionable.

In this talk I show that Karttunen's conditions follow from a basic principle of the trivalent approach: truth-values that determine the result of a binary operator inhibit presuppositions in its other operand. The resulting generalized filtering principle avoids the proviso problem. It also disallows modification of presuppositions, which has been assumed to be necessary in order to derive so-called "conditional presuppositions" in cases like:

  (3) If Theo is a scuba diver, he will bring his wet suit.

By contrast to (1), from (3) it is possible to infer that Theo has a wet suit *if he is a scuba diver*. In agreement with Mandelkern 2016, I will argue that it is doubtful if such conditional inferences should be characterized as presuppositions. Like the proviso problem, this expectation of classical approaches results from their "proposition-orientated" filtering. Instead, I will reintroduce motivations for Karttunen's filtering paradigm, which takes presupposition triggers as the locus where presuppositions may be inhibited by their local context.

January, 30th, 2020, Nadine Bade

Obligatory Additives as a Window into Exhaustivity Implicatures

In this paper, I present a new analysis of the obligatory insertion of additive markers such as the English presupposition trigger "too". Specifically, I argue that additive particles such as "too" and "also" are only obligatory when they need to block or cancel unwanted exhaustivity implicatures. I will discuss the results of a series of studies which support this view and speak against the analysis of obligatory additivity based on the principle Maximize Presupposition (MP) (Heim 1991). Based on more recent findings, I will propose a revised version of Bade’s (2016) analysis of obligatory "too" working with Obligatory Implicatures (OI).

I argue that the empirical picture arising has important theoretical implications beyond the obligatory insertion of additive markers. The data suggest that both MP and OI are limited in their empirical scope. Not all instances of obligatory presupposition triggers track the special inferences which are the result of MP (Percus 2006, Sauerland 2008). I propose that MP accounts only for the insertion of presupposition triggers which have well defined alternatives. Additive particles belong to the set of triggers which serve as an indicator of where exhaustivity is mandatory. This set of triggers is characterized by the lack of lexical alternatives. The bi-division of triggers argued for further supports and stresses the relevance of structurally defined alternatives (Katzir 2007, Fox&Katzir 2011, Katzir 2014), and gives further insights into where EXH occurs.

January, 23rd, 2020, Andrés Soria Ruiz

Evaluative adjectives and action-guidance

In this work I offer a semantics for evaluative adjectives. Evaluative adjectives include moral, aesthetic and personal taste predicates. They are characterized by the fact that they carry an action-guiding inference: a speaker's S utterance of 'φ-ing is good' invites the inference that the speaker will orient her actions towards φ-ing. I decompose this inference into a semantic entailment that the speaker prefers φ-ing (against certain contextually-determined alternatives) and a pragmatic inference that the speaker will act upon her avowed preferences.

January, 9th, 2020, Vincent Lostanlen

Do Androids Dream of Semantic Bleeps?

Machine listening, a subfield of electrical engineering, aims at detecting and classifying acoustic events from natural soundscapes.

Over the past two decades or so, the application scope of machine listening has progressively expanded beyond the realm of automatic speech recognition.

Indeed, it now encompasses all sorts of sounds, including music performance, animal vocalizations, and urban noise pollution.

Such expansion raises a methodological question regarding the essence of machine listening algorithms: can they be regarded as functional models of the human auditory system? and if so, do they extract some form of "meaning" behind auditory events?

Some scholars have suggested that the relative failure of statistical models for content-based music recommendation is evidence of a "semantic gap" between low-level audio features and more abstract concepts, such as genre (cf. Aucouturier).

Others have pointed out that "because music is first and foremost a psychological construct, there can be no externally defined truth" (cf. Wiggins), and that a significant amount of music information relies on subtle syntactic, rather than semantic, mechanisms.

In this talk, I review the current state of the debate surrounding the theoretical framework of music semantics under the light of recent advances in machine listening technology.

First, I demonstrate that musical timbre is tied to playing technique expressivity in ideophonic ways: pizzicato, tremolo, flatterzunge, and so forth.

Secondly, I discuss how human imitations of birdsong span across multiple levels of abstraction: from linguistic description to verbatim copy, via mnemonics, onomatopoeia, whistling, and instrumental decoy.

Thirdly, I show how the design of auditory stimuli by "minimal pairs", originating in music psychology, opens new avenues of research towards self-supervised machine listening.

Examples of such self-supervised tasks are: inferring whether an audio recording is being played forward or backward; whether two melodies are heard in the same tune; and whether two rhythmic patterns fall on the same downbeat.

Concurrently, these findings stress the need for rethinking the relationships between sound and meaning, through coordinated investigation from humanities and experimental sciences.

December 17th, 2019, Chiara Branchini

Syntactic marking in Italian Sign Language: the role of hand selection

I will present the preliminary results of a research that aims at investigating the role of hand selection in the use of classifiers (predicative verbal classifiers and classifiers acting as proforms of nominal referents) in clarifying the arguments’ syntactic roles in a LIS sentence.

December 12th, 2019, Thomas Geissmann (animal communication)

Primate calls and the evolution of singing and music

Loud calls share ancestral sound patterns and functions, and probably occur in all Old World monkeys and apes. They also occur in some New World monkeys and wet-nosed primates, and probably occurred in the last common ancestor of all living primates. Singing and duetting occurs in two groups of the apes: Gibbons (Hylobtidae) and humans (Homo). A phylogenetic study suggests that singing developed twice independently in apes (i.e. in gibbons and humans), but in each case derived from loud calls. Furthermore, the study allows to reconstruct the original contexts and functions of human singing, dancing and, ultimately, music.

December 12th, 2019, Dan Lassiter

Trivalent semantics for indicative conditionals: Motivations and challenges

In 1936 the great probabilist Bruno de Finetti proposed a trivalent, truth-functional theory of the semantics of indicative conditionals and showed that it coheres nicely with a broadly Bayesian perspective on conditional reasoning. The trivalent semantics is much simpler than its main rivals; it provides a simple resolution of Lewis- and Hájek-style triviality arguments about probabilities of conditionals; and it makes available a purely semantic treatment of conditional restriction of modals, quantifiers, and other operators that undercuts much of the motivation for the Lewis/Kratzer restrictor theory that is dominant in linguistics. Some major outstanding issues include framing a general theory of the pragmatics of assertion for trivalent conditionals, and providing a complete and empirically adequate account of conjunctions and disjunctions of conditionals as well as embedded and nested conditionals.

November 28th, 2019, Yimei Xiang

Higher-order readings of WH-questions

In most cases, the proper answer to a WH-question names an entity in the set denoted by the WH-complement. However, Spector (2007, 2008) finds out that the proper answer to a modalized question can be an elided existential quantifier or a Boolean disjunction, as in (1). Moreover,

Xiang (2016) shows that questions with a collective predicate like (2) admit answers naming a Boolean conjunction. These facts show that WH-questions admit higher-order readings, in which the WH-phrases range over sets of higher-order meanings.

(1) Which books does John have to read?

     The French novels or the Russian novels, the choice is up to him.

(2) Which kids formed a team?

     The two boys (formed a team) and the two girls (formed a team).

Strikingly, if the WH-complement is singular-marked or numeral-modified, the proper answer to this question cannot be an elided conjunction, as seen with multi-choice answers to mention-some questions.

(3) What can we use (as a textbook) for this class?

     Book A {or, and} Book B, the choice is up to you.

(4) Which textbook can we use for this class?

     Book A {or, #and} Book B, the choice is up to you.

This talk has two goals: (i) to determine what higher-order meanings should and should not be ruled into the quantification domain of a WH-phrase; and (ii) to derive the two types (i.e., regular and conjunction-rejecting) of higher-order readings compositionally and explain their distributional constraints.

November 21st, 2019, Mathias Sablé-Meyer

The laws of mental geometry in human and non-human primates

One long-standing hypothesis about the uniqueness of human beings is the ability to manipulate a symbolic language of thought that can handle recursive representations. From this faculty others can emerge, in particular language, theory of mind, mathematics and music. This research project explores the extent to which geometry similarly relies on such a capacity. By comparing the performance of different populations on carefully crafted geometrical tasks, we

outline what could serve as primitives in a language of thought for geometry.

For instance, the mental representation of a square appears to be immediately enriched in humans with symbolic properties such as its equal-length sides and parallelisms, while performance of non-human primates suggest that these representations are absent. While this outlines some primitives of a language of thought for geometry, it says nothing of the control structures, or recursive abilities, of the language. We therefore moved to a second set of

experiments where we model the complexity of complex shapes (e.g. a circle made of squares) solely using the complexity of simple shapes (in this case, circle and square), thereby showing that any good candidate for a language of thought of geometry ought to have a few control structures, notably repetition, concatenation and embedding.

November 14th, 2019, Diogo Almeida

When an error is an error is an error is an error: the processing of subject-verb agreement violations

Subject-verb agreement is a (potentially) long-distance dependency. Since subjects and verbs may be separated at a (potentially) linearly unbounded and unpredictable fashion, it poses a problem for the parser, as information about the subject needs to remain accessible throughout the ongoing processing stream until the verb is encountered, so agreement can be checked. Selective failures of subject-verb agreement have been reported both in production (e.g., Bock & Miller, 1991) and perception (Wagers, Lau & Phillips, 2009), when the subject is separated from the verb by an intervening Noun Phrase (e.g. "The key to the cabinets was/*were on the table"), directly implicating memory constraints in the real time processing of agreement information. In these situations, an erroneous agreement marker in the verb is produced or perceived as more acceptable than it should when it matches the agreement feature of the intervening NP, even when the latter is in a structurally unavailable position to govern agreement. This kind of selective fallibility is generally discussed either in terms of models that posit a fuzzy encoding of sentence structure in working memory, or rather models that posit retrieval failures of subject information from an otherwise faithful encoding of sentential structure. Similarly, these models implicitly or explicitly assume that agreement errors are uniform across agreement features, both in terms of values and types. However, most of the data in this literature deals only with Number agreement, and have a tendency to rely on structures in which the subject is morphologically unmarked and the intervenor is morphologically marked. On the basis of a series of experiments in Modern Standard Arabic (MSA), a language that allows Number and Gender agreement on the verb, I will argue that agreement errors are sensitive to the nature and the value of the agreement features of the subject, and discuss potential consequences for syntactic and processing accounts of subject verb agreement.

November, 7th, 2019, Claire Pelofi

Cortical tracking of high-level properties of music

Music, just as language, is a highly complex set of information, which contains structural layers ranging from low-level acoustical features (pitch, timber), to high-level syntactic or semantic information. The way these features are intertwined and encoded into the brain to ultimately convey musical meaning and emotion remains largely unknown. Electrophysiological studies have been focusing on ERP analysis, in which a specific feature is probed through numerous repetition. The time-locked neural response is then averaged to shed light on the encoding of the information. Albeit informative, this technique poorly reflect ecological conditions of music listening. Recent research have developed new tools specifically designed to track ongoing information in the neural signal using canonical correlation analysis (CCA) and temporal response function (mTRF). These two methods are based on ridge regression and matrix linear transforms which allows to measure the mapping of stimulus features into the neural signal. My research is dedicated to study the encoding of high-level musical properties, taking advantage of these new signal analysis techniques and working with neurophysiological signals collected from EEG, MEG or ECog. This line of research tackles questions such as the encoding of regularities in music, the neural trace of tension and release dynamics and quantify the effect of cultural learning.

November 7th, 2019, Olivier Morin

The Color Game: Studying the evolution of an artificial language in an online multiplayer gaming app

The Color Game (live from spring 2018 to spring 2019) was a multiplayer app whose participants were invited to play a series of referential communication games. One player attempted to communicate one colour, the target, which the other player had to pick among a set of four. Players communicated through a set of black and white symbols that carried vague chromatic connotations, insufficient to allow beginners to communicate with high accuracy. The app allowed any player to play with whomever they wished. It was downloaded by three thousand players and played more than three hundred thousand times. Before releasing the app, we pre-registered six studies, five of which were carried out. This talk will present two of them, and in so doing introduce the Color Game dataset. The first study attempted to explain how symbols would gain information, their referential meaning progressing from vague to precise. We predicted that this process would be driven by the increasingly selective use of the pictures carrying the most precise connotations, and the gradual abandonment of less informative ones. Even though players do make greater use of high-information symbols over time (confirming our preregistered prediction) we found that the amount and kind of information carried by particular symbols overwhelmingly depended on individual players' usage, and increased gradually as these players gained experience, without the need for any selective process. The second study tested a predicted link between pragmatic relevance and cultural attraction in the evolution of the Color Game's referential conventions. Starting from the well-known fact that certain colours are more readily named and recognised than others, we predicted, and found, that the interpretation of Color Game messages would be biased towards these salient colours. Salient colours were both easier as referents and harder as distractors: a slient target colour made success more likely, while a salient distractor made it less likely. In keeping with our predictions derived from cultural attraction theory, salient colours also served as "anchors" in the evolution of the game's codes: symbol-colour mappings are more stable for salient colours.

October 31st, 2019, Léo Migotti & Léo Zaradzki

Walk-denoting music: refining music semantics

Recent attempts have tried to account for the meaning of music thanks to a formal model inspired by formal semantics, i.e. that the meaning of X is the set of situations in which X is true (Schlenker 2017, 2019). Because music is by nature very different from language, though, the semantics of music is necessarily far more abstract than that of natural languages. In order to check whether such an abstract model resists people's actual intuitions, we consider a very specific kind of music that can trigger strong inferences about virtual 'walking events' (in which a character is walking), for these inferences tend to be extremely strong and consistent. First, we will try to identify the musical parameters involved in the triggering of such inferences, and show that they might not be restricted to the ones identified by Schlenker. Second, we will show that new conditions on the rules of the model need to be stated. Specifically, we offer an alternative system which is based on correspondance relations between the mereological structure of music and the mereological structure of its denotation. We will argue that these mereological structures are better respresented as headed trees, and that semantically heads and non-heads are to be treated in different ways. This then leads us to explore the ultimate question of compositionality in music.

October 24th, 2019, Charlotte Hauser

Relative clauses in French Sign Language – an eye-tracking study

Subject relatives (SRs) are easier to process than object relatives (ORs) in many spoken languages: French[1], English [2], German[3], Korean[4] or Japanese[5]. However, relative clauses (RCs) are not solely found in this modality, various sign languages also display relativization strategies[6][7]. Few experimental works exist on sign language processing[8][9] and none on RC processing. One major challenge of such study comes from the adaptation of any audio-visual paradigm to a visual-only one. We ran an eye-tracking experiment on spoken French (SF), using the traditional visual world paradigm with the exception that stimuli consisted of texts, where the sentence was appearing phrase by phrase. We replicated results previously obtained through the audio-visual design for SF[10]. We then ran the same experiment with French Sign Language (LSF) signers using LSF stimuli. Participants (30 native/nearly native LSF speakers and 21 SF native speakers) watch a sentence (Table 1) while viewing a pair of pictures with the same three characters, each performing different actions (Fig.1). Stimuli were written for SF (Fig.2) and signed for LSF (Fig.3). Participants had to find the correct picture corresponding to the sentence. One of the pictures was only compatible with a SR interpretation, the other one only with an OR interpretation. Using maximal linear mixed models[11][12], we found a SR advantage in LSF and SF, which is in line with the robust SR advantage found in many spoken languages.

October 17th, 2019, Nadine Bade

New data on the nature of competition between indefinites and definites

I will report novel experimental data on the status of anti-presuppositions (or ‘presuppositional implicatures’), which are generally seen as inferences resulting from reasoning over presuppositional alternatives, based on comparisons between definites and indefinites. The results pose challenges for recent proposals in the theoretical literature that these inferences, just like quantity-based implicatures under a grammatical approach (Fox 2007, Chierchia et al. 2012), are derived via a (on certain accounts mandatory) mechanism of exhaustification (Magri 2009, Marty 2017, Elliott & Sauerland 2018) and should thus be generally treated on par with quantity implicatures. In particular, we find that anti-uniqueness inferences for indefinites pattern differently from both implicatures and presuppositions across experimental tasks.Our results show that in absence of an overt competitor that matches the presupposition, indefinite and definite are indistinguishable when it comes their ability to refer to unique discourse objects. Based on these findings, we hypothesize that the competition between the definite may only be evoked by certain readings of the indefinite. In the presence of a competing picture with a unique object a quantificational meaning may be at play, which does enter competition with the definite, and is utilized in order to resolve the forced choice; whereas in the case where only a unique referent is overtly given a referential reading of the indefinite may be the most relevant, for which the definite alternative does not play a role.

October 10th, 2019, WooJin Chung

Intervention in Deontic Reasoning

The ‘if p, ought p’ problem, famously known as Zvolenszky’s puzzle (Zvolenszky 2002), questions how we can systematically distinguish non-defective cases of ‘if p, ought p’ from defective ones. The issue is that certain facts stand and fall together, and for this reason it is not trivial to set up a coherent modal background which would let us predict the non-defective instances of the ‘if p, ought p’ problem to be non-vacuously true. I propose that a counterfactual-based formulation of deontic necessity that implements intervention, jointly with the assumption that indicative conditionals facilitate backtracking, offers a deterministic account of the ‘if p, ought p’ problem. I also present empirical evidence in favor of an interventionist approach to counterfactuals as opposed to similarity-based theories, at least in the domain of deontic reasoning.

July, 25th, 2019, Mel Cosentino (animal communication)

I beg you pardon? Intra and interspecific communication of wild of harbour porpoises

Harbour porpoises are difficult to observe at sea due to their small size and cryptic behaviour, but they are well-suited for passive acoustic monitoring (PAM), using recording devices coupled with a sound detector and a classifier. Unlike other toothed whales, harbour porpoises produce only highly stereotyped clicks, which other species use for echolocation purposes: seeing the surroundings using sounds. We now know, however, that the behaviour of porpoises can be deduced from the pattern of click trains (a series of clicks emitted as a unit), including echolocation and communication. My research is focused on developing a stand-alone tool (PorCCA) to study harbour porpoise behaviour in the wild, by identifying, extracting, and analysing click train patterns. The tool includes a high-accuracy classifier, as well as a series of algorithms to estimate the minimum number of animals present in a given acoustic event, the likely behaviour, and the presence of calves. Additionally, I am studying a unique case of potential inter-specific communication between harbour porpoises and a solitary common dolphin, Kylie, who has lived in a restricted area in West Scotland for almost two decades. Since 2004 she has been frequently seen spending time with a harbour porpoise. I have analysed acoustic recordings from 3 encounters of Kylie and a harbour porpoise and the results indicate that Kylie changes her vocalisations, producing porpoise-like clicks, especially when travelling in close proximity to the porpoise.

July, 11th, 2019, Laurent Bonnasse-Gahot

Efficient communication in written and performed music

Since its inception, Shannon's information theory has attracted interest for the study of language and music. Recently, a wide range of converging studies have shown how efficient communication pervades language, from phonetics to syntax. Efficient principles imply that more resources should be assigned to highly informative items. For instance, average information content was shown to be a better predictor of word length than frequency, revisiting one of the famous Zipf's law. However, in spite of the success of the efficient communication framework in the study of language and speech, very little work has investigated its relevance in the analysis of music. Here, we examine the organization of harmonic information in two large corpora of Western music, one made of MIDI files directly sequenced from scores, and the other made of MIDI recordings of live performances of highly skilled piano players. We show that there is a clear positive relationship between (contextual) information content of harmonic sequences and two essential musical properties, namely duration and loudness: the more unexpected an harmonic event is, the longer and the louder it is.

July, 4th, 2019, Valentina Gliozzi

The emerging lexicon: towards a neural network account

I will present some preliminary results on a neural network model of the emergence of taxonomic and semantic relations in the infant lexicon. In the model thematic relations are learned through Hebbian learning, and taxonomic relations derive from conceptual reorganization. The model aims to be psychologically plausible, and to overcome drawbacks of previous neural network models.

June, 24th, 2019, Angelika Kratzer

Situations, truthmakers, and the interpretation of questions

June, 20th, 2019, Dave Barner

Alternatives and exhaustification in language acquisition

Though children begin to use logical connectives and quantifiers earlier in acquisition, studies in both linguistics and psychology have documented surprising failures in children's interpretation of expressions. Early accounts, beginning with Piaget, ascribed these failures to children's still burgeoning semantic and conceptual representations, arguing that children acquire ever more powerful logical resources as they development and acquire language. But more recent accounts, drawing on a Gricean divide between semantics and pragmatics, have argued that certain of these failures might not reflect semantic incompetence, but instead changes in children's pragmatic reasoning abilities. In particular, early studies argued that children might be more "logical" than adults, perhaps because of difficulties with Gricean reasoning, or theory of mind. In this talk, I investigate this question, and argue that

neither pragmatic incompetence nor conceptual/semantic change can explain children's behaviors, and that instead children's judgments stem from difficulties with "access to alternatives".

First, I consider the case study of scalar implicature, and show that children when children hear an utterance like the one in (1) they fail to compute a scalar implicature like in (3) because they are unable to spontaneously generate the stronger alternative scale mate in (2). But when scalar alternatives are provided contextually or are "unique" alternatives, children no longer struggle with implicatures. I show that children easily compute "ad hoc" implicatures and ignorance implicatures (where all relevant alternatives are provided in the original utterances), as well as inferences that exhibit similar computational structure, like mutual exclusivity. Also, I show that children's problems cannot be ascribed to difficulties with epistemic (theory of mind) reasoning, ruling out the idea that their problems are related to understanding other minds and intentions.

(1) I ate some of the cake

(2) I ate all of the cake

(3) I ate some (but not all) of the cake

Having concluded that a non-Gricean model of implicature is mandated by previous data, I then attempt to assess some such approaches, including the multiple exhaustification hypothesis, which predicts, inter alia, that children should compute free choice inferences, but also, given a lack of access to alternatives (like conjunction), that they should treat disjunctive statements like (4) conjunctively.

(4) Every girl ate a cookie or a cake

I present a direct replication of this finding, and then show two experiments which argue that children's behaviors are not attributable to conjunctive meanings, but instead stem from problems of infelicity in the original experiments. I then present a follow-up experiment, using an entirely different paradigm, which again finds very few conjunctive children, but then provides independent evidence that these children's judgments are not consistent over time. I argue that, in fact, children generally don't compute any implicatures whatsoever, except for (1) free choice, and (2) the distributive inference. I note that because the multiple exhaustification account deploys the same mechanism for free choice inferences and disjunctive statements like in (4), something is amiss, and that an alternative account is needed. In particular, based on novel experimental evidence from children, I explore the possibility an inclusion account like Bar-Lev & Fox (2017) or Santorio & Romoli (2017).

May, 29th, 2019, Milica Denic

Donkey anaphora in non-monotonic environments: An experimental investigation (joint work with Yasutada Sudo)

Donkey pronouns in sentences such as (1) are ambiguous between an existential (a) and a universal (b) interpretation. Which reading is preferred depends on the quantifier: universally quantified sentences like (1) favor the universal reading, while existentially quantified sentences strongly favor the existential reading.

(1) Every farmer who owns a donkey loves it.

(a) Every farmer who owns a donkey loves some of his donkeys.

(b) Every farmer who owns a donkey loves all of his donkeys.

When it comes to donkey pronouns in the scope of non-monotonic quantifiers, which readings are available and what their order of preferences is remain debated questions, and different theories are tailored for different intuitions (Kanazawa 1994, Champollion et al. 2018). We address these questions experimentally for two non-monotonic quantifiers: exactly 3 and all but one. Surprisingly, we find that these two quantifiers differ with respect to which readings they allow. We discuss how our results bear on the theories of donkey anaphora ambiguities, and what factors can explain the differences between exactly 3 and all but one.

May, 23rd, 2019, Igor Yanovich

Approximate Bayesian Computation for learning about the linguistic past

Many models of complex processes are so complex themselves that precise statistical inference about them becomes infeasible. Approximate Bayesian Computation (ABC) is a technique that makes approximate inference possible for an important class of models where it is too hard to deal with likelihood directly, but comparatively easy to simulate from the model. This class includes many reasonable models of population histories, such as the histories of languages and of the genes of the people who speak them.

In the talk, I introduce a specific historical-linguistic problem: the initial stages of the great Bantu expansion across the Subsaharan Africa. The linguistic history of that family featured extensive cross-branch borrowing of both lexical and grammatical material, so the simple family-tree model cannot adequately describe what happened to the Bantu. Including borrowing into the model, however, vastly increases its complexity, making direct inference infeasible. I introduce the framework of Approximate Bayesian Computation, and explain how it can be used to address the Bantu problem, presenting pilot results obtained jointly with Andrea Benazzo, Silvia Ghirotto, and Patricia Santos.

May, 22nd, 2019, Cristiano Chesi

A formal perspective on linear order, intervention and complexity (in object clefts)

In this talk I will present a formal approach to syntactic derivation which is Minimalist (Chomsky 1995, Stabler 1997, Collins & Stabler 2016) and processing-friendly (Chesi 2015).

According to the standard Minimalist perspective, linearization is a matter of Sensory-Motor interface, therefore independent from the core syntactic (recursive) engine (Fox e Pesetsky 2005).

Here I will concentrate on asymmetries that suggest the opposite to be plausible: the linear order has a clear impact on the hierarchical structure and it could be considered as strongly dependent from hierarchy (as in Kayne 1994).

Focusing on performance data gathered from processing tasks based on non-local dependencies in which a clear intervention configuration is attested, namely object-cleft constructions (e.g. “it was [the barber]i [that [the banker] avoided _ i at the party]”, Gordon et al 2001, Warren & Gibson 2005 a.o.), we will discuss how none of the “standard models” is currently able to explain the whole set of asymmetries revealed.

The result of two new experiments (one on-line, eyetracking, the other off-line, acceptability judgments), targeting restricted pronouns in OC configuration (“it was [the/you barbers]i [that [the/you bankers] avoided _ i at the party]”) will also be discussed (Chesi & Canal 2019).

On the one hand, it will be clear that processing-driven theories here discussed, in the current form, are not sufficient to explain how subtle morphosyntactic variants of this configuration have such a high impact on processing; on the other, competence-based theories are not sufficient to predict any on-line processing effect. I will then opt for a fully explicit computational model that is be able to predict, moment-by-moment, the difficulties observed both in on-line processing and off-line acceptability judgments.

May, 9th, 2019, Mora Maldonado

Person of interest: Learnability and naturalness of person systems

Person systems---typically exemplified in pronoun paradigms (e.g. me, you, us)---describe how languages categorize entities as a function of their role in speech context (i.e., speaker(s), addressee(s), other(s)). Like other linguistic category systems (e.g. color and kinship terms), not all ways of partitioning the person space into different forms are equally likely cross-linguistically. Indeed, while some partitions are extremely frequent, others are very rare or do not occur at all (Cysouw 2003).

Morpho-semantic approaches to person systems have aimed to provide an inventory of personfeatures that generates all and only the attested partitions (Harley & Ritter 2002, Harbour 2016, Ackema & Neelman 2018, among others). One potential problem with all these accounts is that the typological data they rely on is rather weak: not only the sample of languages is quite small, but also there are often inconsistencies in the way paradigms are classified.

To make up for this issue, we aim to provide a method for investigating person systems experimentally. Such an approach allow us not only to extend the presently sparse typological data, but also to test whether typologically attested partitions are more natural and easier to learn than unattested ones.

In this talk, we will present a series of artificial language learning experiments where we test whether typological frequency correlates with learnability of person paradigms.

We will start by focusing on first person systems (e.g., ‘I’ and ‘we’ in English), and test the general predictions of theories that posit a universal set of features to capture this space. Our results provide the first experimental evidence for feature-based theories of person systems. We will then present some ongoing research where we take a similar approach to investigate potential asymmetries between the first, second and third person(s).

May, 2nd, 2019, Catherine Hobaiter (animal communication)

Human apes: human use and understanding of great ape gesture

All great apes, including humans, employ a rich repertoire of vocalizations, facial expressions, and gestures to communicate. Great ape gestural repertoires are particularly elaborate, with chimpanzees and bonobos employing over 80 different gesture types intentionally: that is towards a recipient and with a specific goal in mind. African great apes, that is chimpanzees, bonobos, and gorillas, share almost 80% of their gestural repertoires; but what happened to this gestural inheritance in the human lineage? In biological terms chimpanzees and bonobos are more closely related to us than they are to a gorilla. I will discuss two new approaches to this question that employ ape field methods to exploring human behaviour. With focal follows of young early-verbal children we show that production of a substantial proportion of the 'ape' repertoire in human infants, and with play-back of ape behaviour to human adults, we show that comprehension of the ape repertoire appears to have been retained, even in fully language-using individuals. I will discuss the implications of this for comparisons of communication across ape species, and for our understanding of the evolution of human language.

May, 2nd, 2019, Diego Feinmann

Domain Restriction: The Problem of the Variable Location Revisited

Stanley (2000, 2002; Stanley and Szabó, 2000) put forward several arguments in support of his Nominal Restriction Theory (NRT), the view that domain variables are introduced by nouns, and not by determiners as classically stipulated (e.g. Fintel, 1994). This initiated a heated debate between NRT supporters and advocates of the classical account, a debate which eventually came to a deadlock in view of the conflicting evidence. In this talk, I shall examine (and question) the arguments that have been offered in support of, as well as against, NRT. In addition, I shall present an argument, based on newly gathered evidence, that strongly indicates that the classical account is on the right track.

April, 25th, 2019, Janek Guerrini and Léo Migotti

1) Musical gestures in the typology of linguistic inferences

Léo Migotti and Janek Guerrini

Schlenker 2018a argues that pro- and post-speech manual gestures make it possible to replicate with iconic means most of the inferential typology of language, including implicatures, presuppositions, supplements and 'homogeneity inferences' characteristic of definite plurals. Tieu et al. 2018 confirm these generalizations with experimental means, and extend them to hybrids of written words and visual animations (replacing gestures). Independently, it has been argued (Schlenker 2018b) that music can trigger semantic inferences. Putting these paradigms together, we suggest that i) onomatopoeias (vocal gestures) and, more importantly, ii) musical stimuli can replace pro- and post-speech gestures, and trigger the same four inferential types. In other words, musical stimuli can behave as 'musical gestures'. Crucially, musical gestures seem to trigger typically linguistic inferences through specifically musical features like consonances, dissonances, and cadences.

2) Does vowel quality interact with iconic lengthening?

Janek Guerrini

In spoken language it is possible to modulate the length of a given vowel in order to convey a strengthened meaning, e.g.|long talk| < |looong talk|. This very same lengthening is not felicitous for adjectives like short (* “shooort”). For this reason, the lengthening of “large”-type adjectives like “long” is usually held to be purely iconic (Schlenker 2016, Fuchs et al 2019), i.e. the result of a direct mapping from, e.g., the length of the talk to the length of the word “long”. However, it must be noted that for adjectives like “teeny”, the lengthening is possible. Consequently, we argue that to account for iconic modulation of vowel length it is necessary to consider, alongside ‘pure’ iconicity, the back/front opposition of vowels, one of the most robust phenomena linked to sound symbolism. We submit that two mechanisms underlie modulation of vowel length: i) ‘Pure’ iconicity, mapping the length (or number of replications) of the vowel directly onto the size of the object of which the adjective is predicated, thus applying to ‘large’-type words only. ii) Intensification of the vowel symbolism, placing restrictions on the lengthenable vowel requiring the vowel type (back/ front) to ‘match’ with the semantic direction of the adjective (‘large’-type/’small’-type respectively). We present two pilot studies that test acceptability judgements on scalar adjectives whose stress vowel has been lengthened. On a more speculative note, we argue that iconic lengthening and vowel symbolism are far more similar than was previously thought. We argue that they fulfill the same iconicity criterion and that their different functioning arises as a consequence of contingent meaning-relevance in a language. More specifically, in most dialects of English, vowel length is not meaning-relevant, whereas vowel quality is: thus one can arbitrarily lengthen “big”, but not arbitrarily push its stress-vowel to backness: “bag” is just another word.

April, 18th, 2019, Alia Martin

Infants’ understanding of communicative interactions

Communication is a powerful way to share our thoughts with one another. How do infants begin to identify communicative interactions and recognize when communication will be successful? I will present evidence that infants use information about the communicative signal, the social context, and the shared knowledge between speaker and listener, to evaluate third party communicative interactions. Infants' understanding of how information is transferred between people through communication can provide insight into language learning, as well as the early development of theory of mind.

April, 4th, 2019, Hans Rott

Difference-making conditionals and the Relevant Ramsey Test

This talk explores conditionals expressing that the antecedent makes a difference for the consequent. It employs a 'relevantised' version of the Ramsey Test for conditionals in the context of the classical theory of belief revision due to Alchourrón, Gärdenfors and Makinson (1985). The idea of this test is that the antecedent is relevant to the consequent in the following sense: a conditional is accepted just in case the consequent is accepted if the belief state is revised by the antecedent _and_ fails to be accepted if the belief state is revised by the antecedent's negation. The connective thus defined violates almost all of the traditional principles of conditional logic, but it obeys an interesting logic of its own.

The talk also offers the logic of an alternative version, the 'Dependent Ramsey Test' according to which a conditional is accepted just in case the consequent is accepted if the belief state is revised by the antecedent _and_ is rejected (e.g., its negation is accepted) if the belief state is revised by the antecedent's negation.

March, 21st, 2019, Yasutada Sudo

Plurality inferences as quantity implicatures

Plural nouns typically give rise to 'plurality inferences', e.g. "Andrew wrote papers" implies that Andrew wrote multiple papers. Plurality inferences are not always present, e.g. "Andrew did not write papers". There are three types of approaches in the literature: (i) the scalar implicature approach (Spector 2007, Zweig 2009, Ivlieva 2013, Mayr 2015), (ii) the ambiguity approach (Farkas & de Swart 2010, Grimm 2013, Martí 2018), and (iii) the antipresupposition approach (Sauerland 2003, Sauerland et al. 2005). I propose a new scalar implicature account. The scalar implicature account assumes that the plural is semantically number-neutral, and the plurality inference arises as a scalar implicature in competition with the singular. However, the scalar implicature computation is not straightforward, given that pairs like "Andrew wrote papers" and "Andrew wrote a paper" would be truth-conditionally identical. Different versions of the scalar implicature account make use of different truth-conditional asymmetries, e.g. non-global levels of meaning, strengthened meaning, etc. I propose instead that the plurality inference can be derived as a global-level quantity implicature based on non-truth-conditional aspect of the meaning. Specifically, the singular and plural sentences differ in anaphoric possibilities: the plural sentence introduces a discourse referent that ranges over singular or plural entities, while the singular sentence introduces a discourse referent that only ranges over singular entities. Based on this asymmetry, a quantity implicature is derived that the discourse referent is meant to only range over plural entities. This analysis requires no embedded implicatures or higher-order implicatures. I will formalize this idea in update semantics, and demonstrate that it makes correct predictions about negative sentences and quantified sentences.

February, 28th, 2019, Amir Anvari

Indexical Shift and the Ban Against Illeism

Third-person reference to the speaker or the addressee, or illeism, is unacceptable. A plausible explanation, here dubbed the Ban Against Illeism (BAI), is that all else equal indexical pronouns are preferable to third-person noun phrases. BAI makes non-trivial predictions for languages that allow indexical shift. In such languages indexical pronouns in indirect discourse may take their value from the actual or reported context of speech. The first prediction is that indexical shift can bleed BAI: third-person reference to the actual speaker/addressee should be acceptable in a context-shifted environment. The second prediction is that indexical shift can feed BAI: third-person reference to the reported speaker/addressee should be unacceptable in a context-shifted environment. Building on data from Farsi, and assuming the operator-based approach to indexical shift (Anand & Nevins 2004), I will argue that the first prediction is born out without further ado. The second prediction, although also empirically accurate in Farsi, raises two challenges. Empirically, there are attested counter-examples to it in Zazaki (Anand 2006). Theoretically, it requires the assumption that BAI is insensitive to de re/de se distinction. Building on Sharvit 2010, I address the latter problem while making a conjecture regarding the former. Time permitting, I will take a step back and compare the operator-based approach with the binding-based approach to indexical shift (Schlenker 1999). I will argue that although (supplemented with independently needed stipulations) the binding-based approach captures the basic paradigm, there is one data point that is problematic for it but not its alternative.

February, 21st, 2019, Maria Aloni

Modal inferences of the 3rd kind: the case of free choice

In free choice inferences conjunctive meanings are derived from disjunctive sentences contrary to the prescriptions of classical logic:

(1) You may eat pizza or pasta => You may eat pizza and you may eat pasta.

Free choice inferences present a challenge to the canonical divide between semantics and pragmatics. Although derivable by conversational principles they lack other properties of canonical pragmatic inferences: they are often non-cancellable, they are sometimes embeddable (e.g., under universal quantifiers) and their processing time has been shown to equal that of literal interpretations. In this sense they are neither purely semantics nor purely pragmatics, they are inferences of the 3d kind.

In the talk I will present a state-based semantics where such 3rd kind inferences are derived by allowing pragmatic principles intrude in the recursive process of meaning composition. Contrary to most existing accounts where free choice inferences are viewed as special cases of Quantity implicatures, the relevant pragmatic principle in our logic-based approach will be a version of Grice's Maxim of Quality.

February, 14th, 2019, Toshitaka Suzuki (animal communication)

Avian linguistics? Referentiality and compositionality in bird calls

In human speech, words often convey independent meanings and syntax allows combining multiple words into more complex, compositional expressions. In contrast, animal communication signals have typically been considered as motivational: vocalizations merely reflect emotion or arousal of signalers and do not provide compositional messages even when multiple units are combined. However, recent field studies have challenged this assumption by showing that several species of birds and nonhuman primates may be able not only to assign independent meanings to acoustically discrete vocalizations, but also to combine these signals into higher lexical sequences. In this talk, I introduce my recent studies on vocal communication in a small bird species, the Japanese tit (Parus minor). Japanese tits produce acoustically discrete calls in a variety of contexts, such as when encountering a predator and when facilitating group cohesion. Field experiments have revealed that these calls may not merely reflect arousal of signalers, but also convey information about external referents, such as the presence of a predator. In addition, these birds are able to combine meaningful calls into higher structured sequences according to an ordering rule. Playback experiments suggest that receiver tits are able to use an ordering rule to extract compound meanings from call sequences even if these sequences are composed of novel combinations of calls. These findings may demonstrate a new parallel between bird calls and human language, opening new avenues for exploring the origins and evolution of linguistic capabilities in nonhuman animals.

February, 14th, 2019, Elitzur A. Bar-Asher

A new proposal for the interpretation of sentences with the reciprocal anaphors

It is well-known that sentences with expressions such as each-another in English, exad et-haSeni in Hebrew, and un-l'autre in French can have various the interpretations with different logical strengths (in terms of entailments). This state-of-affairs has repeatedly raised the following questions, using Dougherty's (1974: 18-19) words: "How is a specific input linked to a specific output? That is, what is the rule of semantic interpretation for each other sentences?… how a specific interpretation (or range of interpretations) is assigned to an arbitrary sentence?"

Naturally, there are at least two competing directions for answering these questions:

Strongest Meaning Hypothesis (SMH) (Dalrymple et al. 1998; Sabato & Winter 2012): The meaning of the reciprocal sentences varies from one sentence to another and is taken from a small inventory of meanings. It is possible to predict a context-sensitive meaning of every reciprocal sentence: in a given context, a sentence takes the strongest meaning that is consistent with known facts about the specific context. (cf. Mari 2014).

The Unspecified Constructions Hypothesis (UCH): The relevant constructions are ‘unspecified constructions’, described by the following definition: expressions used in relations between two (defined) sets (or more) without specifying which set occupies which position. Accordingly, the basic meaning of these constructions has a much weaker meaning than what SMH argues for, because it only necessitates that each member of the set stands in a single relation to another member. Although this is necessarily true for all sentences with these expressions, these are not sufficient conditions to capture only true sentences in many cases.

The paper will take the second approach and will aim at proposing a mechanism for how the semantics of such sentences interacts with their context.

February, 7th, 2019, Cornelia Loos

Responding to negative assertions: A typological perspective on yes and no in German Sign Language

Response particle systems vary across languages in terms of the number of particles and in terms of their discourse functions. Where some languages have two particles (like English yes and no), others have three (like German ja, nein and doch). Traditional accounts of such response systems make a distinction between truth-based and polarity-based systems (see, e.g., Pope 1976, Jones 1999). In truth-based systems, yes-type answers confirm the truth of the antecedent proposition (1bi, 2bii) while no-type answers disconfirm, i.e. reject, it (1bii, 2bi). In polarity-based systems, on the other hand, response particles signal the polarity of the response clause as either positive (yes-type particle, 1bi, 2bi) or negative (no-type particle, 1bii, 2bii). As (1-2) illustrate for English, languages may also employ both response systems and use no to either reject the truth of a proposition (1aii) or signal the negative polarity of the response (2bii). Languages with a tripartite system, like German, often have a dedicated response particle for rejecting negative propositions (scenario 2bi), although other types of dedicated particles exist as well (Roelofsen & Farkas 2015).

(1) a. Anna smokes.

 b. i. Yes (=She does). ii. No (=She doesn’t)

(2) a. Anna doesn’t smoke.

 b. i. Yes/ ? No (=She does). ii. ?? Yes/No (=She doesn’t)

When it comes to the visual-gestural modality, very little is known about the inventory of response particles in sign languages (but see Gonzalez et al. on ASL), including their role in signaling truth vs. polarity. Sign languages are of particular interest here since they have multiple articulatory channels available which may simultaneously encode truth and polarity. The present study provides experimental data from a production experiment with 24 native signers of DGS, which investigated responses to positive and negative assertions in DGS. The study shows that DGS favors a truth-based over a polarity-based strategy but also exhibits some modality-specific response strategies that combine truth and polarity.

January, 31st, 2019, Hazel Pearson

Constraints on Counteridenticals

I will consider a particular class of attitude report known as ‘counteridenticals’ – reports involving predicates such as imagine, pretend and wish where the attitude holder puts herself in the shoes of someone else, as in the following example.

1. Yesterday, Margaret was imagining that she was Helen. She imagined that *she* was going to marry Mr Wilcox.

The bolded pronoun is ambiguous. On the counterfactual-self reading, it is Helen who is going to marry Mr Wilcox, whereas on the belief-self reading, it is Margaret. I identify two

constraints on the availability of these readings: (i) the counterfactual-self reading is not

available when the verb is anchored to the belief worlds of the attitude holder; (ii) the belief-

self reading is not available when the pronoun or anaphor is unambiguously read de se (eg

PRO, shifted indexicals, long-distance reflexives) and the verb is anchored to the counterfactual worlds of the attitude holder (cf Pearson 2018).

These findings have implications for the analysis of de se construals: I show that approaches

that treat de se as a special case of de re struggle to account for the data (eg Reinhart 1990).

By contrast, they can receive an elegant analysis if we assume that the expressive resources of the grammar include a mechanism of ‘de se binding’, as originally proposed in Chierchia

(1990). The data thus provide a means to resolve a long-standing debate in linguistics and

philosophy about the proper analysis of de se phenomena.

January, 24th, 2019, Maribel Romero

Counterfactual biscuit conditionals and the role of temporal and mood morphology

(joint work with Eva Csipak)

*Hypothetical* conditionals convey that the truth of the consequent clause is dependent on the truth of the antecedent clause. They may appear in indicative form, as in (1), or in subjunctive/counterfactual form, as in (2), with different morphological marking appearing in the two: An extra layer of past tense –labeled 'fake' tense– and (if available) subjunctive mood typically characterize the counterfactual form. This fake tense has received a modal remoteness analysis (Iatridou 2000, Schulz 2014) and a temporal remoteness analysis (Arregui 2009, Gronn & Stechow 2009, Romero 2017).

(1) If John had a hangover yesterday, he was in bed.

(2) If John had had a hangover yesterday, he would have been in bed.

*Biscuit* conditionals, instead, intuitively assert the truth of the consequent regardless of the truth or falsity of the antecedent, as in (3). Concentrating on the indicative versions, a prominent line of analysis argues that hypothetical conditionals like (1) and biscuit conditionals like (3) have exactly the same syntactic structure and modal interpretation and differ only in the pragmatic inference process (Franke 2009, 2015, Lauer 2015, Csipak 2018, Biezma & Goebel t.a.). Such analysis is facilitated by the fact that the two forms share the same verbal morphology.

(3) There are biscuits on the sideboard if you want them. [Austin 1956]

However, when we come to counterfactual biscuit conditionals, verbal morphology sets them apart from their hypothetical counterparts: The consequent clause of the hypothetical counterfactual (2) has richer morphology –namely, would have been in (2)– than that of the hypothetical biscuit –were in (4). The question arises, why two different forms are needed if the intuited hypothetical vs. biscuit interpretation is simply a matter of pragmatic inferencing.

(4) There were biscuits on the sideboard, had you wanted them / if you had wanted them.

The present talk combines the temporal remoteness approach to counterfactual hypothetical conditionals and the aforementioned pragmatic approach to biscuit conditionals to propose a (tentative) explanation of the morphology observed in biscuit conditionals.

December, 13th, 2018, Moshe Bar-Lev

Homogeneity and the distributive-collective distinction

This talk focuses on two questions in the semantics of plural predication:

(1) Does plural predication give rise to `specified' (i.e., distributive or collective) meanings or to `underspecified' ones (compatible with both distributive and collective situations)?

(2) What is the source of variation between non-distributive predicates with respect to Homogeneity (Križ 2015)?

Examining question (1), I argue that both specified and underspecified meanings should be derivable (following Schwarzschild 1991; Heim 1994), and observe that predicates differ in their Specification properties (whether they give rise to specified or underspecified meanings). I further claim that there is a correlation between the Specification properties of predicates and their Homogeneity properties, which calls for a unified perspective on questions (1)-(2). I propose such a perspective based on a novel trivalent semantics for Link's star operator (following but departing from Schwarzschild 1994) together with a relativization of that operator to `covers' (Schwarzschild 1991, 1994; Heim 1994).

December, 6th, 2018, Lara Mantovan

A matter of contrast: Exploring irony in Italian Sign Language (LIS)

Verbal irony is a matter of contrast: an ironic remark conveys a meaning that is typically the opposite of what is literally said. If the addressee fails to recognize the intended meaning, there is a risk of misunderstanding and communication failure (Kreuz et al. 1999). To make sure that the addressee can detect irony and correctly interpret the utterance, the ironist might use irony markers, i.e. meta-communicative cues that alert the addressee that the utterance requires an ironic interpretation (Attardo 2000). As far as we know, despite the vast literature on the use of irony markers in spoken languages, no research has been devoted at identifying irony markers in those languages that exploit the visual modality to convey meaning, i.e., sign languages.

To start filling this gap, we administered to four Deaf native Italian Sign Language (LIS) signers a Discourse Completion Task to obtain a semi-spontaneous elicitation of 10 minimal pairs of ironic/literal remarks expressing either compliments or criticisms. The analysis of this corpus revealed that: i) sentence meaning is expressed manually through the polarity of the evaluative lexical sign, ii) signer's attitude is expressed non-manually through mouth-corners up and down, iii) ironic remarks systematically show a prolonged articulation, and iv) irony might be further signaled by non-obligatory non-manual and manual cues.

November, 29th, 2018, Zoe Schlueter

Reasoning in a second language does not make you more rational when it comes to loss aversion

Research on cognitive reasoning biases has shown that people are predictably irrational in their decision-making (Kahneman 2011). However, some recent work suggests that they are less prone to classic cognitive biases in their second language (L2) than in their first language (L1) (Costa et al. 2014, Keysar et al. 2012). This “foreign language effect” has been argued to show that decision-making heuristics are less dominant in L2 than L1 reasoning and has been attributed to a lower emotionality associated with the L2 (Costa et al. 2014). However, an alternative explanation is that the materials in these experiments are susceptible to multiple interpretations and a decision that is irrational under one interpretation is perfectly rational under another interpretation (as shown in monolinguals by Mandel 2013). Therefore, the difference between L1 and L2 speakers might be due to subtle differences between native and non-native interpretations affected by proficiency. We present data from two experiments examining the effect of proficiency on loss aversion bias in native Spanish speakers at different levels of English proficiency. We do not replicate the finding that L2 users in general are “more rational” in their decision making than L1 speakers, but the results indicate that loss aversion bias is affected by proficiency in the L2. Moreover, our results suggest that high proficiency L2 speakers behave like L1 speakers in respect to loss aversion, both with materials susceptible to pragmatic enrichment and with materials inducing exact interpretations.

November, 22nd, 2018, Jacopo Romoli

More free choice and more inclusion: An experimental investigation of free choice in non-monotonic environments

(based on joint work with Nicole Gotzner and Paolo Santorio)

Disjunctions in the scope of possibility modals give rise to a conjunctive inference, generally labeled ‘free choice’ (Kamp 1973). That is, a sentence like Iris can take Spanish or Calculus typically suggests that she can choose between the two. An approach which has become prominent in the literature derives free choice as a kind of scalar implicature (Fox 2007; Klinedinst 2007; Franke 2011 a.o.). In this talk, I focus on the predictions of two main type of accounts within this approach, with the goal of investigating what is the best implicature-generating algorithm for capturing free choice and related data points. The first is based on a standard algorithm for computing implicatures, which proceeds by negating, ‘excluding,’ alternatives to a sentence and adding the information so obtained to the assertion. Accounts in this vein have been successful in capturing free choice and related phenomena. Recent experimental findings involving negative quantifiers have, however, challenged exclusion approaches (Chemla 2009). On these grounds, Bar-Lev & Fox (2017) have proposed a novel theory of free choice, which can account for those data. The main goal of this talk is to explore the exclusion vs. inclusion debate further, by looking at another case in which the predictions of the two accounts diverge, involving disjunction in the scope of non-monotonic quantifiers. The crucial case is sentences like Exactly one girl cannot take Spanish or Calculus, and its potential free choice reading suggesting that one girl cannot take either Spanish or Calculus and all of the others can choose between the two. I report on an inferential task experiment testing this case, building on Chemla 2009; Chemla & Spector 2011 and Gotzner & Romoli 2017. In our results, we find evidence for this free choice reading. As we discuss, this is challenging for the exclusion implicature theories of free choice, but in line with inclusion accounts like Bar-Lev & Fox 2017. This case constitutes, therefore, a further argument for inclusion accounts.

November, 15th, 2018, James Hampton

Ducks lay eggs – is the acceptance of minority characteristic generics dependent on implicit gender differentiation

Generic statements assert default properties of a kind. They reflect the relevant features of our concepts and are considered by people generally true of the entire class despite the existence of counterexamples (e.g. birds fly). We report three experiments which explore the factors that lead to the acceptance of generic statements. In particular we examine whether properties that relate to gender differences (lions have manes, or ducks lay eggs) are more likely to be accepted than matched statements that refer to an arbitrary subclass unrelated to gender.

November, 7th, 2018, Michael Wagner

Toward a Bestiary of the Intonational Tunes of English

What is the inventory of tunes of North American English? What do particular tunes contribute to the pragmatic and semantic import of an utterance? How reliably are certain conversational goals and intentions associated with the use of particular tunes? While English intonation is well-studied, the answers to these questions still remain preliminary. We present the results of scripted experiments that complement existing knowledge by providing some data on what tunes speakers use to accomplish particular conversational goals, and how likely particular choices are. This research complements studies of the meaning and form of individual contours, which often does not explore the alternative prosodic means to achieve a certain conversational goal; it also complements more exploratory research based on speech corpora, which offer a rich field for exploring which contours are generally out there, but since the context often underdetermines the real intentions of the speaker, they make it hard to come to firm conclusions with respect to the contribution of particular tunes.

Our studies focus on three types of conversational goals, the goal to contradict (‘Intended Contradiction’), to imply something indirectly (‘Intended Implication’), or to express incredulity (‘Intended Incredulity’). We looked at these three intents since their expression has been linked in the prior literature with the use of three particular rising contours: the Contradiction Contour (Liberman & Sag, 1974; Ladd, 1980; Ward & Hirschberg, 1985; Goodhue & Wagner 2018)), the Rise-Fall-rise Contour (Ward & Hirschberg, 1985; Constant, 2012; Wagner, 2012; ), and the incredulity contour (Hirschberg & Ward, 1992).

Our results show that participants indeed use the expected contours more frequently than others to achieve the respective conversational goals---except that they almost never used the Incredulity Contour. To convey incredulity, speakers almost always chose the Polar Question Rise (Pierrehumbert & Hirschberg, 1990, Bartels, 1999; Truckenbrodt 2012). In Contradictions, there was more variability in the choice of intonational tune than with the other two intents. When speakers did not use the Contradiction Contour, they often contradicted the interlocutor using a Declarative Fall with Polarity Focus, or a hitherto undescribed falling contour, which we label the Presumption Contour. Our results also show an interesting interaction between choice of tune and focus prominence (Goodhue & Wagner 2016; cf. Schlöder 2018). We discuss the challenge such interactions pose for Rooth's alternatives theory of focus, and how one might go about addressing it.

October, 25th, 2018, Marie-Christine Meyer

L+H* imposes restrictions on the quantifier domain of L% and H%

In this talk I will propose a compositional semantic account of the intonational patterns L+H*LH% and L+H*LL%, which are often used in contrastively negated statements (e.g., Mary DIDN'T go to PARIS. Implied: Mary went somewhere else). The overarching goal of the talk is to highlight three crucial components of my analysis and the general insights they offer for the theory of (scalar) implicature and the semantics of intonation. First, the symmetry problem that inevitably arises with (focused) negation. Second, the compositionality problem. This concerns the interaction between the final fall (LL%) vs. final rise (LH%) and the pitch accent in giving rise to the implicated meaning component (in our example: that Mary went somewhere else/that the speaker is uncertain whether she went somewhere else). Third, the ambiguity problem. Whatever semantic denotation we assign to L% (and H%), these morphemes seem often vacuous except for their discourse structuring role.

October, 11th, 2018, Robert Pasternak

Parts and wholes in the domain of attitudes

One area of inquiry in event semantics that has borne a great deal of fruit there, but that has generally been put aside in Davidsonian treatments of propositional attitudes, is how the part-whole structure of events can affect entailments and acceptability in a variety of constructions. In this talk I will discuss two types of evidence suggesting that attitude states have non-trivial part-whole structures (beyond the obvious case of temporal part-whole relations). First, I will argue that measurements of intensity track the part-whole structure of states of desiderative attitudes like desire and regret; put simply, a more intense desire state is "bigger" along some (possibly abstract) dimension than another, less intense desire state. My evidence comes from a variety of measurement constructions that impose requirements relating measurement to part-whole structure---including pseudopartitives (Krifka 1989, Schwarzschild 2006) and nominal and verbal comparatives (Wellwood et al. 2012, Wellwood 2015)---all of which can be used to measure the intensity of attitude states. Second, I will demonstrate the availability of non-distributive readings of belief ascriptions. Both of these facts will be analyzed within the confines of a broadly Hintikkan semantics for attitudes.

October, 4th, 2018, Roger Levy

Gender bias in preferred linguistic descriptions for expected events

Language production and comprehension reflects rapid integration of  diverse information sources. Stereotypes---implicit mental associations among concepts---can influence event expectations and  thus bias preferred linguistic descriptions. Do linguistic biases transparently reflect event expectations, or can the mappings between event expectations and preferred linguistic descriptions themselves be biased? Here we demonstrate that the mapping from event expectations  to preferred linguistic descriptions can indeed be biased. In particular, when making pronominal references to individuals whose gender is not known or determined, expectations that the individual might be female manifest in feminine-pronoun preferences at a lower rate than expectations that the individual might be male manifest in masculine-pronoun preferences.  In a large-scale experimental study during the 2016 U.S. Presidential campaign, this pattern was strongly manifested in comprehension and production of pronouns referring to the next U.S. President. This pattern also generalizes more broadly in pronominal production preferences to a wider variety of contexts. These findings constrain the structure of quantitatively precise  theories of pragmatics. Additionally, these findings exemplify how quantitative, experimentally controlled psycholinguistic investigation can identify biases in mappings from events to preferred linguistic descriptions, revealing and deepening our understanding of implicit cognition.

September, 27th, 2018, Emile Enguehard

Connectedness as a constraint on exhaustification

The so-called grammatical theory of scalar implicatures argues that these inferences arise from the application of a covert operator exh, which has the capability to be embedded under other operators. However, in many examples, not all conceivable positions that exh could occupy are attested. Various approaches based on logical strength and monotonicity have been proposed to justify the limited distribution of exh; they are mostly based on a comparison between possible parses, and considerations of monotonicity (e.g., the Strongest Meaning Hypothesis). We propose a new constraint based instead on “connectedness”, ruling out parses because of inherent problems. Connectedness is a sister notion of monotonicity, which has been recruited to explain certain lexical restrictions on nouns, adjectives and more recently quantifiers; we propose here that connectedness could play a similar role at the level of propositional meanings.

June, 27th, 2018, Thierry Aubin (animal communication)

Adaptation des processus de codage acoustique à l’environnement : études expérimentales sur le chant des oiseaux

Plus que tout autre vertébré, les oiseaux ont développé les communications acoustiques. Les échanges de sons jouent en effet un rôle très important dans l’accomplissement de leurs fonctions vitales : se nourrir, se reproduire, se situer dans l’espace, s’identifier, avertir d’un danger, indiquer son état émotionnel etc... Pour communiquer efficacement, les oiseaux ont été amenés à adapter leurs signaux acoustiques aux milieux dans lesquels ils vivent et donc ne communiquent pas de la même façon dans une plaine, une forêt…ou une ville. Certaines espèces utilisent des signaux très complexes, les chants, constitués de séquences de sons organisés suivant une syntaxe bien spécifique et qui véhiculent simultanément différentes informations : identité de l’espèce, identité géographique, identité sexuelle, identité individuelle, état émotionnel. Ces différents niveaux d’information sont codés. Par des analyses acoustiques et des expériences de diffusions à des individus de chants dont on a modifié certains paramètres on démontre que ces codes intègrent les propriétés acoustiques de l’environnement pour adapter leur fonction : l’information publique (codage des identités spécifiques et sexuelles) repose sur des codes du signal résistants à une propagation longue distance et l’information privée (codage des identités géographiques, individuelles et état émotionnel) sur des codes peu résistants à cette propagation.

June, 27th, 2018, Dominique Sportiche

De de re ipsa

A study of the behavior of obligatorily or non obligatorily controlled PRO in locally non attitudinal contexts yields the conclusion PRO and its antecedent must (in simple cases) be contextually intensionally equivalent, the de re ipsa requirement. I will suggest that so called mandatory de te readings, and de se readings found in obligatory control constructions under attitude control predicates can be seen as subcases of this de re ipsa requirement.

In turn (if time permits), starting from this behavior of PRO and comparing it with that or pronouns and reflexives suggest a binding theory in which the binding relation is the sharing (or non sharing) not of reference but of content, i.e. of at least some relevant description returning the same individual for some relevant attitude holder having the participants of this relation within its scope. This provides a different look at some binding puzzles discussed in Heim (1994), and Sharvit (2010) (and possibly in Heim, 2007 and Sharvit, 2014).

June, 20th, 2018, Philipp Koralus & Vincent Wang

Explaining success and failure of human reasoning via questions and arbitrary objects: The erotetic theory

The capacity to reason is central to all advanced human endeavors. Pushed to its limits, this capacity makes possible science and modern civilization. Equally remarkably, we are subject to systematic failures of reasoning in all domains, ranging from simple puzzles to medical diagnoses and legal decisions. These striking failures have been widely experimentally documented, and their discovery has yielded more than one Nobel Prize over the past decades. The reality of both successes and failures of thinking poses a dual challenge for our understanding of the human mind. To meet this challenge, we will need an account of our minds that can make sense of how failures arise, but that also explains how it is possible for us to systematically reason correctly. So far, there is a lack of formally precise theories the can make sense of success and failure for reasoning with quantifiers like “some” and “all” and relations in anything like full generality. Formally precise theories are a prerequisite for connecting the study of reasoning properly with the study of language. We extend the erotetic theory of reasoning for this purpose. The erotetic theory of reasoning holds that the relationship between questions and answers is central to our thinking. We seek to answer questions as directly as possible, which can lead us astray. However, if we raise enough questions, to the point of reaching erotetic equilibrium, fallacies are eliminated. Crucially, we model universal claims in terms of arbitrary objects. This allows us to maintain an attractive view on which the activity of reasoning, from the point of view of the reasoner, is based on content rather than mental formulas. We will provide an introduction to the erotetic theory and present some new data that raises interesting questions about the relationship between reasoning and language.

June, 13th, 2018, Jeremy Goodman

Verbalism

Superman is Clark, and Lois knows that Superman flies, but she doesn’t know that Clark flies. This looks like a counterexample to Leibniz’s law. According to verbalism, this appearance is misleading because the above sentence involves equivocation on “know”. The proposition that Superman flies is the proposition that Clark flies, and Lois knows this proposition in some ways but not in others. The fact that rational people can be confused or unsure about which things are which corresponds to a distinctive kind of context-sensitivity in propositional attitude verbs. We show how to reconcile this ubiquitous contextualism with a strong logic of propositional attitudes, and also how it allows us to reconcile coarse-grained theories of propositions with failures of logical omniscience.

May, 30th, 2018, Ofra Magidor

Counting and copredication (joint work with David Liebesman)

Copredication is the phenomenon exhibited by sentences such as ‘Three heavy books are informative’. The puzzle is that such sentences ascribe two properties that, at least prima facie, cannot be jointly instantiated (e.g. being informative - a property of informational objects; and being heavy – a property of physical objects).

The puzzle of copredication has been addressed by various theorists (Pustejovsky (1995), Luo (2012), Asher (2011), Gotham (2017)) by what we call ‘the dual object strategy’. Different ways of implementing the strategy differ considerably in detail, but the rough idea is to let nouns involved in copredication (‘book’) denote an entity of some kind of “dual” semantic type, one that in some sense combines the two aspects (e.g. being physical and being informational), and which allows, perhaps after some further footwork, the two conflicting predicates (‘heavy’ and ‘informative’) to apply.

However, dual-object theories face a serious challenge from issues concerning counting and individuation. Consider a bookshelf containing three physical volumes, each binding together War and Peace and Anna Karenina. While there seems to be a reading on which we individuate books as physical objects (generating the true reading of ‘Three books are on the shelf’), and one on which we individuate books as informational objects (generating the true reading of ‘Two books are on the shelf), we seem to get no reading on which books are individuated as if it were a sums of both (there is no true reading of ‘Six books are on the shelf’).

In the first part of this talk, I will discuss the most detailed attempt to resolve this tension, by Mathew Gotham (Gotham (2017)). We argue that despite its ingenuity, Gotham’s theory ultimately fails. And while this failure does not prove that any alternative theory cannot succeed, it is illustrative of the challenges of devising an adequate semantics for counting sentence within the dual object strategy.

In the second part of the paper, I will briefly sketch discuss an entirely different solution to this problem, based on our own account of copredication (Liebesman and Magidor (2017)). On this alternative account, nouns such as ‘book’, are restricted (even in copredicational contexts) to either picking out physical objects or to picking out informational objects. This gives an entirely straightforward account of the counting data but still leaves the puzzle of copredication. We argue, however, that the puzzle of copredication has been driven by false metaphysical assumptions: e.g. that physical books cannot be informative or that informational books cannot be on shelves. Once we realise the properties in question are not really incompatible, data about both copredication and counting can be easily accommodated.

May, 16th, 2018, Kyle Blumberg

Counterfactual Attitudes and Paired Propositions

I raise a problem for standard semantics for attitude verbs. The problem I raise involves so-called "counterfactual" attitude verbs, such as ‘wish’. In short, the trouble is this: there are true attitude reports ‘S wishes that P’ but there is no suitable referent for the term ‘that P’. The problematic reports illustrate that the content of a subject’s wish is intimately related to the content of their beliefs. I capture this fact by moving to a framework in which ‘wish’ relates subjects to sets of pairs of worlds, or paired propositions, rather than—as is standardly assumed—sets of worlds. I argue that this two-dimensional framework can also be used to solve other puzzles as well, e.g. the "existence problem" involving indefinites under non-doxastics discussed by Schoubye (2013). I end with some outstanding problems for the account, as well as some suggestions for future research.

May, 9th, 2018, Daniel Goodhue

Must p is felicitous only if p is not known

In recent work (von Fintel & Gillies 2010, Matthewson 2015, Lassiter 2016, Mandelkern 2016), epistemic modals have been claimed to have felicity conditions that require the evidence for the prejacent to be indirect. In contrast, I argue that epistemic modals have felicity conditions that require that the prejacent is not known as claimed in Giannakidou & Mari 2016. New linguistic data is produced in support of this position. The proposed account is argued to explain the new evidence better than accounts that rely on indirectness. The evidence in favor of this account also militates in favor of a weak semantics for must p. In light of these findings, future prospects are explored. In particular, I suggest that this proposal paves the way for the felicity conditions of epistemic must to be derived as a conversational implicature. Furthermore, I demonstrate that a purported counterexample to the proposal, must p statements in the conclusions of deductions, is a problem for indirectness accounts as well, and I suggest a way forward.

May, 2nd, 2018, Diane Lillo-Martin

American Sign Language pronouns and their acquisition

Pointing is a ubiquitous activity that humans engage in from avery young age. For this reason, the analyses of index finger pointing in sign languages and the development of these points in signing children has been a matter of great interest and some controversy. Traditionally, pointing is interpreted as pronominal in sign languages such as American Sign Language(ASL); however, differences are noted between pointing signs and pronouns,which have led some to consider the signs to be a mixture of linguistic and gestural. Here we report new data on asymmetries in the emergence of pointingin Deaf signing children, hearing bimodal bilingual children, and hear ingnon-signers. These data show that children do not treat all points the same,and they contribute to reconsideration of the analysis of pointing and its role in the pronominal system of ASL.

April, 25th, 2018, Katie Collier (animal communication)

Call concatenation in wild meerkats

Repertoire size, frequently determined by the number of discrete call types, has been used to assess vocal complexity in animals. However, species can also increase their communicative complexity by using graded signals or by combining individual calls. Animal call sequences can be divided into two main categories, each subdivided into two classes: repetitions, with either an unlimited or finite number of iterations of the same call type, and mixed call combinations, composed of two or more graded or discrete call types. Social contexts involve a wide range of behaviours and, unlike predation contexts, can be associated with both positive and negative emotions. Therefore, interactions linked to social contexts may place additional demands on an animal's communicative system and lead to the use of call combinations. We systematically documented call combinations produced by wild meerkats, Suricata suricatta, a highly social carnivore, in social contexts in their natural habitat. We observed 12 distinct call combinations belonging to all four classes of combination, produced in all the observed behavioural contexts. Four combinations were each produced in a specific context whereas the remaining eight were produced in several contexts, albeit in different proportions. The broad use of combinations suggests that they represent a non-negligible part of meerkat social communication and that they can be used in flexible ways across various behavioural contexts. Comparison with combinations produced in predation contexts indicated that social call combinations are more varied in number of classes and structural complexity than the former, perhaps due to the greater variety of social contexts. However, in meerkats, combinations of functionally referential calls have been documented in predation but not social contexts, suggesting that both social and predation pressures may play a role in the evolution of combinatoriality in animal communication.

April, 18th, 2018, Raphaëlle Malassis (animal communication)

Artificial Grammar Learning in Baboons

A current dominant hypothesis on the evolution of syntactic abilities propose that the processing of supra-regular grammars isa unique human capacity. In support of this hypothesis, artificial grammar learning studies conducted so far do not provide unambiguous demonstration of this capacity in a non-human species. In this talk, I will argue that these previous failures can be attributed to a bias in these species towards the exploitation of local regularities and difficulties for processing more distant relationships, rather than an inability to master supra-regular grammars. I will present a series of experiments conducted in baboons and humans during myPhD at the station of primatology of the CNRS (Rousset, France), under the direction of Joel Fagot and Arnaud Rey. Our aims were twofold: (1) verifying that the detection of long-distance dependencies is particularly difficult for non-human primates, and (2) assessing baboons’ ability to master supra-regular grammars when they are prevented to exploit local regularities. Overall, our results reveal a stronger continuity in grammar processing capacities within the primate order than previously thought,but also highlight important species differences in memory constraints

April, 4th, 2018, Paul Portner

The Semantics of Action and Comparison in Infinitives and Subjunctives

This talk represents an effort to better understand the semantics of clauses which express notions of action and comparison, such as the following types: (1) Call me tomorrow! (imperative) (2) She would like (for you) to call tomorrow. (comparative infinitive) (3) Il exige que tu partes maintenant. (comparative subjunctive; Mulder 2010) (4) Hafdis knows who to talk to at the party. (wh infinitive; Bhatt 1999) (5) The screwdriver to use is a Phillips. (wh relative; Stanley and Williamson 2001) I focus on two puzzles related to these types of clauses: the nature of the future-orientation in their semantics and the obviation effect (i.e. the preference for a control infinitive rather than a subjunctive or infinitive with an overt subject, when control would be possible). Building on ideas about desire predicates from Heim (1992), about infinitives and control from Farkas (1992), about imperatives from Barker (2012), and about sentence mood from my own work, I propose that non-indicative clauses can contain one of a family of operators introducing related concepts of action and comparison. The analysis allows an unusual but compositional analysis of evaluative attitude predicates like ‘want’ and an appealing pragmatic theory of imperatives. It supports a treatment of subject obviation as a case of Maximize Presupposition and explains why relative futurity is linked to the semantics of comparison and control. To conclude, we might speculate about the relation between the types of non-indicative clauses subject to the analysis, and non-comparative relatives like ECM infinitives, polarity subjunctives, and non-comparative wh infinitivals.

March, 28th, 2018, Roumi Pancheva

Cardinality measures and number marking on nominals

Nominals modified by numerals show variation in number marking, cross-linguistically and even within one and the same language. Ionin and Matushansky (2006) propose that numerals always combine with lexical noun phrases which denote predicates of singularities; plural marking on this view is a form of agreement. Bale et al (2011), however, propose that all numeral modification is restrictive, prohibiting the combination of numerals with semantically singular noun phrases (what they call ‘the strong thesis’). In languages where numerals combine with morphologically singular (bare) noun phrases, this is because nouns are semantically number neutral. In this talk I present evidence from Bulgarian that numerals can combine with semantically singular noun phrases. The argument involves a reanalysis of one type of number inflection for masculine nouns (the ‘count’ form), which has traditionally been considered a form of agreement on otherwise semantically plural nouns. If that argument is correct, the ‘strong thesis’ is not. I end with a suggestion that what underlies variation in number marking is that there are two routes to obtaining cardinality measures.

March, 21st, 2018, Jeremy Kuhn

Negative concord as split scope

'Concord' describes a phenomenon in which a single logical meaning is expressed syntacticallyon multiple lexical items. The canonical example is negative concord, in which multiple negativeexpressions are used, but a single negation is interpreted. In some dialects of English, (1) negatesthe proposition that Ed saw someone. Likewise, in ‘distributive concord,’ multiple words with distributive marking may appear innocently in the same sentence, with a singledistributive meaning, as in (2), from American Sign Language.

(1)  Ed didn’t see nobody.

      ‘Ed didn’t see anybody.’

(2)  BOY EACH(distr) CHOOSE-distr ONE-distr GIRL.

      ‘The boys each chose one girl.’

A standard view is one in which negative concord items do not themselves bear negative meaning, but are in an obligatory syntactic relationship with an operator elsewhere in the sentence (Zeijlstra 2004). For distributive concord, on the other hand, Henderson (2014) and Kuhn and Aristodemo (2017) argue that each distributive marker is semantically interpreted, and that distributive marking on verbs and numerals is a wide-scoping predicate that checks that a plurality of events has been introduced.

Here, we extend an analogous semantic analysis to negative concord. We hypothesize that concord (in all guises) is a split-scope phenomenon, consisting of the dynamic introduction of a discourse referent and a cardinality test on that referent. Scope-taking plus dynamic semantics allows us to consider properties of objects that emerge via interaction with other logical operators. Specifically, we propose that negative concord items introduce a discourse referent (like an existential), but then test that no discourse referent has been introduced in any assignment. These apparently contradictory requirements are licensed with split scope around negation: introduction occurs below negation; the test appears above it.

March, 14th, 2018, Ur Shlonsky

French wh in situ

My talk will consider a number of empirical and conceptual problems in the syntax of French wh in situ. I will attempt to explain why it is excluded in indirect questions, permissible in “strong” islands but degraded in “weak” ones and I will examine some of the intervention effects that constrain its distribution.

March, 7th, 2018, Claudia Poschmann

Local NRCs - Experimental Evidence from German

Generally, it is assumed that non-restrictive relative clauses (NRCs) strongly project and, despite their embedded position, cannot take narrow scope under operators of their host-clause (McCawley 1982, Potts 2005, Arnold 2007, AnderBois et al. 2011, Simons et al. 2011). Schlenker (2013), however, provided examples from English and French in which the NRCis interpreted locally (1a). Interestingly, such local readings are not always readily available.See the contrasts in (1) below. But what makes the difference between (1a) and (1b)?

(1) a. If tomorrow I called the Chair, who in turn called the Dean, then we would be indeep trouble.

b. *If tomorrow I called the Chair, who hated me, then we would be in deep trouble.(Schlenker 2013)

In this talk, I will present the results of three experiments in German testing the availability oflocal readings depending on the clause-type (NRCs, parenthetical, conjunction, postponed matrix clauses), the predicate-type (state/event) and the presence or absence of certain discourse particles (like ”then”). In contexts incompatible with a global interpretation, NRCs were significantly more often accepted as suitable than the corresponding parentheticals and postponed matrix clauses. This strongly confirms the assumption that NRCs can have local interpretations (Schlenker 2013). Moreover, the results suggest that the availability of such embedded readings is dependent on the predicate type of the NRC and the type of the rhetorical relation established between the NRC and its host-clause (coordinating vs. subordinating). In the final part of this talk, I will discuss potential explanations for these findings and make some suggestions for further research on this topic.

February, 14th, 2018, Clemens Mayr

On the relation between the triggering and the projection problem for presuppositions

In this talk I investigate Schlenker’s 2010 view whereby the triggering and the projection problem for presuppositions might be closely related. Following suggestions in the literature, I will argue that presuppositions are simple entailments. However, they need to be entailed by the utterance context because the truth-conditions of the sentence containing them are not fully specified. Otherwise, the sentence would not be assertable in the context. On this picture it turns out that the projection and the triggering problem are indeed two sides of the same coin. This view still necessitates a certain amount of lexical stipulation regarding the presupposition triggers. I suggest that this is unavoidable given a certain flexibility in the triggering and projection patterns. Finally, I attempt to single out a common source for the observed triggering and projection patterns thereby making the requirement on lexical stipulation somewhat more palatable.

February, 7th, 2018, Constant Bonard (music semantics)

Truth-conditional music semantics: two challenges from art criticism

Philippe Schlenker (2017) has proposed an outline for a truth-conditional music semantics. In this talk, I should present two challenges for such a project. The first is that of synonymy and goes as such: (a) If musical meaning is truth-conditional, then there can be different musical utterances which are synonymous. (b) There cannot be different musical utterances that are synonymous. (c) Therefore, musical meaning is not truth-conditional. The defender of a truth-conditional music semantics must reject (a) or (b). I argue that (a) follows from Schlenker’s model as it stands today, and then consider why (b) is plausible. To this end, I present a thesis which has a long history in art criticism – a version of “The Heresy of Paraphrases” (Brooks 1947) – according to which meaningful art don’t allow for paraphrases, unlike, e.g., vernacular, philosophical, or scientific meaningful statements.

The second challenge is that of context dependency and goes as such: (a) If musical meaning is truth-conditional, then two musical utterances made up of the same sounds cannot have different meanings. (b) Two musical utterances made up of the same sounds can have different meanings. (c) Therefore, musical meaning is not truth-conditional. Once again, the defender of a truth-conditional music semantics must reject (a) or (b), I argue that (a) follows from Schlenker’s model as it stands, and then consider why (b) is plausible. To do so, I again adapt a thesis coming from art criticism, sometimes known as the “Pierre Ménard case”, introduced in contemporary aesthetics by Arthur Danto (1981).

My conclusion is not that a truth-conditional semantics is doomed to failure. On the contrary, I believe that taking up these challenges would allow specifying the extent of what can and cannot be dealt with a truth-conditional musical semantics, avoiding potential misunderstandings, and allowing a fruitful dialogue with art criticism, a domain where the meaning of music is richly discussed.

February, 7th, 2018, Mélissa Berthet (animal communication)

When combining creates meaning: the case of titi monkeys

Titi monkeys possess an aerial alarm call, the A-call, that is highly predator-specific, while their

terrestrial alarm, the B-call, tend to be more general and can be used in non-predatory contexts, like when the monkeys are moving near the ground. A- and B-calls can be combined into sequences that may encode both predator type and location, but previous analyses were simplistic and did not investigate whether the information was also processed by listeners.

First, we reassessed the context specificity of B-calls sequences, focussing both on their acoustic and sequential structure. We found that B-calls could be differentiated into context-specific acoustic variants (terrestrial predators vs. ground-related movements) and that call sequences to predators had a more regular sequential structure than ground-related sequences. Second, we investigated whether predator type and location were encoded by the caller and processed by the receivers. We carried out two field experiments (predator presentations and playbacks) and modelled the vocal and behaviour responses of titi monkeys. We found that the proportion of combination of B-calls in a sequence conveyed reliable information about both predator type and location.

Overall, our findings suggest that call combination is a key component of titi monkeys’ alarm system.

January, 31st, 2018, Andreas Haida

Logical reasoning by scalar inference computation

Joint work with Luka Crnicˇ & Yosef Grodzinsky

It seems that logical reasoning is a necessary component of human activities such as science,engineering, and legislature. However, psychological studies found failure rates of up to 80% and more when subjects perform logical reasoning tasks such as forming (in)validity judgments about the arguments in (1) and (2), where only (1) is valid in Aristotelian logic.

• 1. (1)  All desires are sins (p1) and some beliefs are desires (p2)

  2. Therefore: some beliefs are sins (c) (valid)

  3. (2)  All desires are sins (p1) and some beliefs are sins (p3)

  4. Therefore: some beliefs are not desires (d) (invalid)

These findings have often been taken to show that human reasoning is illogical. Other researchers concluded that reasoning is logic-based but also involves non-logical inferential methods such as scalar-inference (SI) computation, which yields the logically invalid inferences in(3).

(3) a. some beliefs are sins (p3) 􏰀 not all beliefs are sins (s1)

b. some beliefs are not desires (d) 􏰀 not all beliefs are not desires (s2)

Recent studies suggest that there is individual variation in the use of SIs. We identify five‘SI profiles’ of logical arguments such as the profile of (2): the premises p1 and p3 validate the conclusion d if and only if the SI of p3 is taken into account and the SI of d is supressed (that is, p1 ∧ p3 ∧ s1 ⇒ d but p1 ∧ p3 ∧ s1 ̸⇒ d ∧ s2). In a validity-rating experiment, we tested for individual variation in the response to the five SI profiles. Our data show, for instance, that there are subjects that systematically accept arguments with the profile of (2) but reject logically invalid arguments that are inert to SI computation. Thus, they compute SIs for premises but not for conclusions. Overall, the data show that a sizable proportion of ‘logical failures’ are in fact successfully performed logical inferences on the basis of sentence meanings that are enriched by SIs. I will discuss how we may explain the existence of different groups of reasoners and the preference to perform SI computations in premises over conclusions.

January, 24th, 2018, Salvador Mascarenhas

Reasoning with disjunctions as a form of hypothesis testing

The idea that human reasoning is best modeled by rational (Bayesian) update procedures resting on a representational system with probability measures has gained great currency in the psychology of reasoning in the 21st century (e.g. Oaksford and Chater, 2007). Conversely, the popularity of research paradigms that approach human reasoning within the mold of (some) deductive system has decreased.

In this talk I argue that elements from both approaches are necessary to understand a diverse class of fallacious inference patterns involving disjunction. More concretely, I give a reconceptualization of the Erotetic Theory of Reasoning of Koralus and Mascarenhas (2013),

a variant of Mental Models Theory (Johnson-Laird, 1983), in terms of Bayesian confirmation theory, where the hypotheses under consideration are determined by disjunctive premises interpreted as questions, along the lines of Inquisitive Semantics (Groenendijk, 2008; Mascarenhas, 2009). I make this case with both purely theoretical arguments and

experimental evidence. The resulting picture of human reasoning is congenial to the view of the human mind as a rational updater (cf. Hohwy, 2013), while demonstrating the need for "old-fashioned" logical elements of model-theoretic approaches to content.

January, 10th, 2018, Keny Chatain

Local contexts for anaphora

Schlenker (2008, 2010) shows a completely systematic incremental procedure to derive local contexts for presupposition. This removes the need for encoding projection behaviour in particular lexical entries, as in Dynamic Semantics. In subsequent unpublished work, Schlenker attempts to apply the same ideas to define an incremental notion of local context relevant to anaphora. In work in progress, I pick up on this project ; I try to account for existential/universal reading of donkey anaphora, in terms of context indeterminacy. I move on to show how the proposal can be used to derive puzzling cases of donkey anaphora with a non-indefinite antecedent: existential readings of definite plurals (Kriz, 2015), derived kind predication (Chierchia, 1998), etc.

December, 13th, 2017, Cornelia Ebert

Temporal sequence and the alignment of gesture and speech

The temporal sequence of verbal expressions as well as the temporal alignment of gesture and speech is decisive for the information status of the involved expressions. It is by now established that, while appositives are generally seen as contributing non-at-issue, sentence-final appositive clauses are much easier to be interpreted at-issue than sentence-medial ones (AnderBois et al. 2014; Koev 2013). Similarly, the temporal synchronization of gesture and speech is not without consequences (cf. Esipova 2017). I will argue (against Schlenker 2016) that while co-speech gestures are not-at-issue by default (Ebert & Ebert 2014; Schlenker 2016), post-speech gestures are more likely to be interpreted at issue. I propose a continuous scale of self-contained gesture interpretation: gestures that have their own time slot are interpreted at issue; the less they are synchronized with speech the more likely it is that they will be interpreted as at issue material (cf. Kendon’s continuum, Kendon 1980).

Furthermore I will discuss the possibility and systematic means to shift information from the non-at-issue dimension to the at-issue dimension and vice versa. I will focus my attention on the relationship between gesture and speech and argue that language provides different means of initiating dimension shifting.

December, 6th, 2017, Lisa Bylinina

On ‘zero’

joint work with Rick Nouwen

Zero is a relatively recent addition to many cultures -- the word 'zero' is first used in English only in the 16th century. However, 'zero' is not uncommon: according Merriam Webster dictionary, 'zero' is more frequent than 'thirteen'. In this talk, we will focus on a prenominal use of 'zero', as in 'I have zero new emails in my inbox'. We show that 'zero' can't be an emphatic variant of 'no' and that giving 'zero' a regular numeral semantics is possible and desirable. We formulate such an analysis and its consequences. In particular, we argue that if zero is indeed the bottom of the numeral scale, then the domain of entities will have to come with a bottom element as well and is a lattice rather than merely a semi-lattice. We explore the effect this will have on the meaning of degree quantifiers in general and on the meaning of bare plural indefinites.

November, 29th, 2017, Yael Sharvit

What negative polarity items reveal about true sentences

Strict negative polarity items (e.g., ‘in years’, ‘until tomorrow') are acceptable in the scope of ‘not’ unless the negative polarity item and some “blocker” are in the scope of the same occurrence of ‘not’, and the negative polarity item is in the scope of the “blocker”. ‘True’ is a “blocker” but ‘think’ is not (as illustrated by the unacceptability of ‘It isn’t true that Mary has had a good friend in years’ vs. the acceptability of ‘I don't think that Mary has had a good friend in years’). We discuss what this fact implies about the syntax and semantics of ‘think S’, ‘know S’ and ‘It is true that S’.

November, 22nd, 2017, Alexandre Cremers

A hybrid approach to the ignorance inference of modified numerals

Modified numerals, such as "at least 3" or "more than 5", tend to trigger ignorance inferences. Geurts&Nouwen (2007) famously argued that these ignorance inferences are stronger with superlative "at least" than with comparative "more than", and proposed a modal denotation for "at least" which semantically encoded the ignorance inference. Since then, competing accounts have been proposed which try to derive all ignorance inferences as implicatures, keeping very simple denotations for "at least" and "more than". 

In this talk, I will first present experimental work showing that (a) there is indeed a difference between "at least" and "more than", but (b) against the predictions of a purely semantic account, the ignorance inference of "at least" is not so strong, and is affected by QUD. Along the way, we also show a contrast between "at least/more than" on the one hand, and "at most/fewer than" on the other hand, and some interesting results with bare numerals.

Although the results support a view of ignorance inferences as implicature, I will argue that purely pragmatic accounts are untenable. I will instead sketch a mixed account in which "at least/at most" behave as modals because of their superlative morphology, but the ignorance inference is ultimately derived as an implicature.

November, 15th, 2017, Heather Burnett & Olivier Bonami

Indexicality, Utility and Conceptual Spaces: Variation and Change in Grammatical Gender in the Debates of the French House of Representatives

In this presentation, we give a new study of the role that social meaning and speaker ideologies play in variation and change in g(rammatical) gender in French. More specifically, we study noms de métiers et de fonctions 'professional nouns' which have the following g-gender assignment pattern: when they are used to refer to socially female individuals, they can have either masculine or feminine g-gender (i.e. le ministre or la ministre for a female minister); whereas, when they are used to refer to socially male individuals, they can have only masculine g-gender (only le ministre for a male minister).

We present (to our knowledge) the first quantitative study of the linguistic and social factors that condition the use of masculine vs feminine g-gender with reference to women, focusing on variation in the transcripts of the debates of the Assemblée Nationale (AN, French House of Representatives). This corpus features intra-speaker variation in g-gender use, as shown by the examples in (1), both said by Jean-Marc Ayrault.

(1)a.Madame le ministre de l’environnement, plus de 6 000 personnes ont défilé, samedi dernier, dans les rues de Nantes, pour protester contre l’autorisation donnée par le Gouvernement à EDF de remblayer la zone humide du Carnet dans l’estuaire de la Loire. (M. Jean-Marc Ayrault, 29/01/1997)

b.Monsieur le président, madame la ministre, mes chers collègues, tout à l’heure, le président Bayrou me reprochait d’avoir dit que nous étions venus pour voter le projet de loi de finances. (M. Jean-Marc Ayrault, 19/12/1997)

The use of grammatical gender in expressions referring to women has been the subject of enormous amounts of prescription and language planning in France and within the Assemblée Nationale itself (see Houdebine 1987, 1998, Burr 2003, Viennot 2014, among others). These efforts can be naturally divided into two phases of activism: First, in 1986, Prime Minister Laurent Fabius legislates the use of feminine grammatical gender and (certain) feminized forms in the AN and similar government institutions. However, we show in our data that this prescription had little to no effect on the speech of the politicians at the time (see also Yaguello 1989, Brick & Wilks 1994 for non-quantitative observations). Second, in 1998, Prime Minister Lionel Jospin issues a statement reiterating Fabius' policy. We show that, unlike 12 years earlier, use of the feminine form (eg. la ministre) successfully replaces use of the masculine form (eg. le ministre) within the space of a year in the AN. This striking difference raises the question: What changed from 1986 to 1998 which allowed the feminine form to take over, possibly aided by (the exact same) language policy?

Our main proposal in this talk is that changes in the use of feminine grammatical gender and differences in the effectiveness of Fabius/Jospin’s language policy are the result of changes in social gender ideologies that occurred in France between the mid 1980s and mid 1990s. To make this claim maximally explicit, we develop a formal model of the relationship between ideological structure and language use and interpretation based on current work in game theoretic pragmatics (particularly Franke 2009 and Frank & Goodman 2012). More specifically, we use Gärdenfors (2000, 2014)'s Conceptual Spaces framework to formalize speaker ideologies and Burnett (2017)'s Social Meaning Game framework to capture the link between ideological structure, social meaning and language use. Using this model, we show that the failure of Fabius' policy and Jospin's subsequent successful use of this policy is predictable from independently motivated assumptions concerning 1) the social meaning of French g-gender (following Livia 2001, McConnell-Ginet 2013), and 2) changes in social discourses surrounding the properties of female politicians associated with theParité ('equal representation') debates in the late 1990s (Ramsay 2003, Scott 2007, Julliard 2012, among others). We therefore conclude that tools from formal semantics and pragmatics can be helpful to understanding both the relationship between social change and linguistic change, and the conditions under which language policies can be effective.

October, 25th, 2017, Amir Anvari

Encoding content: logical force and contextual knowledge

I will provide a unified analysis of three classes of data that, despite the widespread intuition that they have a common source, have so far not been brought under one theoretical roof. Specifically, these are Maximise Presupposition! phenomena (Heim 1991, Percus 2006, a.o.), cases involving presupposed ignorance (Spector & Sudo 2016), and mismatching implicatures (Magri 2009, 2011). My proposal, couched in trivalent semantics, is based on a core condition, which states that an expression must not be used in a context in which it would contextually entail any of its logically non-weaker alternatives: if an alternative is not logically entailed, it must not be contextually entailed either. I will show that this condition, coupled with a suitable projection recipe, captures the paradigm uniformly and makes novel predictions to boot, which I will corroborate.

October, 18th, 2017, Paolo Santorio

Conditional Excluded Middle in Informational Semantics

An empirically adequate semantics for conditionals should vindicate both a principle of Conditional Excluded Middle and the incompatibility of "If p, not q" and "If p, might q". Unfortunately, no existing semantics succeeds in the task. I go on to suggest that the puzzle posed by these two logical requirement is a generalization of Yalcin's well-known puzzle about epistemic contradictions and brings out a basic tension for truth-conditional frameworks. I show that a generalization of Veltman's update semantics, which I call "path semantics", can be used to capture both principles. I conclude by suggesting that path semantics also promises to help with traditional difficulties related to the relation between conditionals and probability.

October, 11th, 2017, Rachel Dudley

Discovering the factivity of "know

"Know" and "think" both express beliefs but differ in their 'factivity'. "Think" can report false beliefs, while the complement of "know" is presupposed to be true. How do children figure out that "know" is factive but "think" isn't. In this talk, I'll report on a series of studies (involving behavioral and corpus methods) designed to pursue this question. Using behavioral tasks, we find that children begin to understand the difference around 3, but that there is a lot of individual variation in when this understanding is demonstrated. One source of this individual variation might be differences in the linguistic experience that children have with "know" and "think". To pursue this hypothesis, we used corpus methods to examine aspects of the input that children receive. We find that direct cues to factivity are sparse: (i) "think" is rarely used in contexts where the complement is false; (ii) "know" is rarely used in contexts where its complement is presupposed. However, we find that "think" and "know" differ greatly in how speakers use them in conversation: (iii) "know" is used to ask or answer questions, whereas "think" is used to make weak assertions. This suggests that noticing the goals of speakers who use the verbs might provide a less noisy signal than observing what speakers presuppose in using the verbs. Finally, I'll introduce a new study in this series (currently in progress) which will directly measure the relationship between input and understanding of the verbs.

October, 4th, 2017, Chris Barker

Negative Polarity as Scope Marking

What is the communicative value of negative polarity? That is, why do so many languages maintain a stock of special indefinites (weak Negative Polarity Items) that occur only in a proper subset of the contexts in which ordinary indefinites can appear? Previous answers include: marking the validity of downward inferences; marking the invalidity of veridical inferences; or triggering strengthening implications. My starting point for exploring a new answer is the fact that an NPI must always take narrow scope with respect to its licensing context. In contrast, ordinary indefinites are notorious for taking wide scope. So whatever else NPIs may do, they at least serve as an utterly reliable signal that an indefinite is taking narrow scope. As also proposed in recent work of Kusumoto and Tancredi, I will show that NPIs are only licensed in contexts in which the wide scope construal of an indefinite fails to entail the narrow scope. In other words, weak NPIs occur only in contexts in which taking narrow scope matters for interpretation. Thus one part of the explanation for the ubiquity and robust stability of negative polarity is that it signals scope relations.

September, 27th, 2017, Brian Buccola

Obligatory irrelevance and the computation of ignorance inferences

(joint work with Andreas Haida, HUJI)

The standard grammatical theory of scalar implicature, as envisioned by Chierchia (2004), Fox (2007), and Chierchia, Fox, and Spector (2012), posits that scalar implicatures are derived in grammar, as a matter semantics, rather than pragmatically, as an implicature rooted in Grice's maxim of quantity. Ignorance inferences, by contrast, e.g. those associated with plain disjunctive sentences, are derived pragmatically, as quantity implicatures. More generally, the standard theory predicts that for any utterance S and any relevant proposition φ which isn't entailed, and whose negation isn't entailed, by S, S gives rise to an inference of speaker ignorance about φ. We argue that this prediction is wrong: it fails to explain the contrast in ignorance inferences associated with "at least" (which obligatorily implies ignorance) vs. more than (which doesn't) (Geurts and Nouwen 2007; Nouwen 2010, 2015). The problem is that, without stipulating restrictions on which propositions are relevant, the theory overgenerates ignorance inferences across the board. We argue that the solution is to close relevance under belief (if φ is relevant, then it's also relevant whether the speaker believes φ). This move has the effect that ignorance inferences, like scalar implicatures, can only be derived in grammar, via a covert belief operator of the sort proposed by Meyer (2013) and discussed further by Fox (2016). The maxim of quantity, we show, then no longer enriches the meaning of an utterance, per se, but rather acts as a filter on what can be relevant in an utterance context. In particular, certain alternatives (of certain utterances) are shown to be incapable of being relevant in any context where the maxim of quantity is active -- a property we dub *obligatory irrelevance*. We argue that obligatory irrelevance provides the key to understanding the contrast in ignorance inferences exhibited by "at least" vs. "more than". We also argue that translating our proposal into neo-Gricean terms, if at all possible, would yield a conceptually less appealing and empirically less adequate theory.

July, 13th, 2017, Matthew Mandelkern

Bounded Modality

To what degree does an epistemic modal claim like ‘It might be raining’ resemble an avowal of ignorance like ‘For all I know, it’s raining’? Progress on this central question in philosophy of language has been made by exploring differences in how constructions like these embed—in particular by exploring their behavior as part of larger constructions like Wittgenstein (1953)’s ‘It might be raining and it’s not’ and Moore (1942)’s ‘It’s raining and I don’t know it’, respectively. A variety of approaches have been developed to account for those differences. All approaches, however, agree that the infelicity of unembedded Moore sentences and unembedded Wittgenstein sentences is to be explained in roughly the same way: these sentences are classically consistent, but commitment to both conjuncts is in some sense incoherent.

In this paper I argue against this consensus. If this consensus were right, then disjunctions of Moore sentences, and disjunctions of Wittgenstein sentences, would be felicitous. This prediction is borne out for Moore sentences, but not for Wittgenstein sentence. This creates a puzzle, since there is decisive reason to think that 'Might p'  is consistent with 'Not p', which suggests, in turn, that their conjunction is consistent. To make sense of this situation, I propose a new theory of epistemic modals and their interaction with embedding operators which makes the interpretation of epistemic modals depends on their local context. I show that this theory accounts for the infelicity of Wittgenstein sentences and their disjunctions. I argue that, moreover, this theory also accounts remarkably well for the subtle behavior of embedded modals across the board. This approach thus sheds new light not only on the semantics of epistemic modals, but also more broadly speaking on the dynamics of information in natural language. My theory, crucially, builds on the symmetric version of Schlenker (2009)’s theory of local contexts. The empirical success of this approach in accounting for the behavior of epistemic modals provides striking confirmation of this kind of approach to local contexts, but with a surprising new perspective: my approach only accounts for the data if local contexts are calculated in an exclusively symmetric manner, contrary to extant approaches to the intrasentential dynamics of information.

July, 5th, 2017, Emeric Henry

Facts, Alternative Facts, and Fact Checking in Times of Post-Truth Politics

How persuasive are “alternative facts,” i.e., false statements by populist politicians, in convincing voters? How effective is fact checking in countervailing alterna- tive facts? We conduct a randomized survey-based experiment to evaluate the impact of alternative facts and fact checking on knowledge, beliefs, and political preferences of voters in the context of the 2017 French presidential election campaign. Marine Le Pen (MLP), the extreme-right candidate who reached the runoff, regularly used alter- native facts in support of her policy proposals, to which mainstream media responded with systematic fact checking. We randomly expose subgroups of a representative sample of 2480 French voting-age population to quotes from MLP and/or real facts. The results are as follows. First, alternative facts used in a logical argument are highly persuasive. Second, fact checking improves factual knowledge of voters, but does not reduce at all voters’ policy conclusions or support for MLP. Third, providing only the true facts backfires by increasing political support for MLP compared to a control group, although to a smaller extent than alternative facts. Finally, heterogeneity of voters with respect to prior voting choices and prior knowledge is important for the effect of treatments on political preferences.

June, 28th, 2017, Alejandrina Cristia

What shapes language (acquisition): Beyond sterile dichotomies

Much ink has been spilled in a debate between "generative/structural" and "usage-based/functional" approaches: Is language acquisition best studied by the description of a hypothetical "language acquisition device", or is it better to assume that linguistic knowledge emerges as a side effect of functional pressures as children try to communicate? In the last 15 years, the very foundations of the debate have been shattered by, on the one hand, a rapprochement to the center from both sides and the emergence of a middle-ground of so-called "statistical" approaches, and, on the other, mounting evidence that certain characteristics deemed necessary by both extreme camps may actually be missing in otherwise functioning linguistic communities. In this talk, I will draw from descriptions of typical and atypical language acquisition research, individual variation studies, and emergent sign languages to discuss (a) why this debate as posited is potentially noxious to the advancement of language acquisition research and, (b) what are the implications of this language acquisition evidence for the language sciences at large.

May, 31st, 2017, Nora Boneh

The Syntax and Semantics of Russian Ditransitive constructions (Nora Boneh & Léa Nash)

In this paper we propose a syntactic analysis of dative DPs in ditransitive constructions in Russian, answering three questions: (I) what semantic roles the indirect object realizes; (II) how it is syntactically ordered with respect to the direct object realizing the theme argument, and (III) how the first two issues are related to the morphological encoding of the indirect object, as a PP or as a morphologically case-marked DP. Addressing first question (II), we show that two kinds of syntactic hierarchies between the two internal arguments of a ditransitive configuration coexist, and that there are two sorts of datives that are hierarchically higher than the theme: those that can reconstruct and those that cannot. We then establish an interpretative correlation between these two types of dative DP, showing that the former is locational and the latter is not, providing the answer to question (I) and elucidating what underlies the morphological similarity, question (III). The interpretative and syntactic differences between scrambled and base-generated high datives lead us to claim that in Russian, dative ditransitives have two distinct underlying structures that are not derivationally related. A scalar approach to event structure enables us to pinpoint the interpretative correlate of each type of dative (locational vs. non-locational) and provides a conceptual argument in favour of a non-synonymy non-derivational approach we pursue here: a path scale encoding event schema cannot be transformed into a different scale based event schema due to movement of the dative DP.

May, 24th, 2017, Charlotte Hauser & Carlo Geraci

Relativization strategies in French Sign Language LSF

Background. Relative structures are one of the key aspects of the syntax of human language because, on a pair with sentential complements and sentential adjuncts, they instantiate sentential embedding. This is one of the reasons why relativization is a well explored area in SL linguistics. Indeed, SLs instantiate the full typology of relative constructions: externally headed (DGS Pfau & Steinbach 2006, LIBRAS Nunes & de Quadros), internally headed (ASL Liddell 1980 and Wilbur & Patschke 1999, i.a.) and correlatives (LIS Cecchetto, Geraci and Zucchi 2009).

Goals. At the empirical level, i) we describe the macroscopic structure of relative constructions in LSF, ii) we document the typology of constituents that can be relativized and iii) we illustrate the properties of the relative markers. At the theoretical level, i) we provide a unified account of the relativization strategies, ii) we provide a strong argument for the raising analysis of relative clauses and iii) we extend the account to cases in which an entire clause is relativized (as in Adyghe, Caponigro & Polinsky 2011). Data are from two native signers of LSF. Relative constructions have been elicited with picture contin- uation tasks and other standard elicitation techniques.

Main strategies. LSF instantiates at least three strategies of relativization: by means of a relative marker (glossed as PI, cf. (1a)), by means of the classifier for person (like in DGS this is only for human referents, cf. (1b)), or via zero-marking (cf. (1c)). The set of nonmanuals includes: eyebrow raising, mouthing (of the relative marker) and upper body orientation towards the location of the head of the relative clause. Spreading is normally limited to the relative marker/head.

(1) a.

b.

c.

rel

IX-1 PREFER VET PI CURE DOG

rel

IX-1 PREFER VET PERSON-CL CURE DOG

rel

IX-1 PREFER VET CURE DOG

‘I prefer the vet that cures the dog.’

‘I prefer the vet that cures the dog.’

‘I prefer the vet that cures the dog.’

Macroscopic structure. These constructions instantiate headed relative clauses with the relative marker delimiting the left periphery of the relative clause. The examples in (1) exclude correlatives because there is no fronting of the relative clause (like in English). Word order facts in object-object relative clauses prove that we are dealing with externally headed relative clauses (cf. (2a)). However, we also found cases where the head remains inside the relative clause (cf. (2b)).

rel

(2) a. IX-1 PREFER MAN [rel PI DOG LICK    ] ‘I prefer the man that the dog is licking.’

rel

b. IX-1 PREFER [rel PI LITTLE GIRL CURE DOG ] ‘I prefer the little girl that pets the dog.’

Microscopic structure. Typologically, languages differ on which constituents can be relativized along the lines defined by Lehmann’s hierarchy (Lehmann, 1988):

Relative clauses in LSF can be constructed over subjects (cf. (1a)), objects (cf. (2a)), adjuncts (cf. (3a)) as well as every intermediate position (not shown here). Interestingly, when the relative marker PI is at the right edge of the relative clause, as in (3b), the whole clause become the head of a relative clause and the interpretation is that of a relative clause on the entire event/situation.

rel

(3) a. IX PREFER TOOTHBRUSH WITH PI GIRL PAINT ‘... toothbrush the girl is painting with.’

Subject > Object > Ind.Object > ... > Adjunct

b. IX-1 PREFER VET CURE DOG PI

‘I prefer situations in which a vet cures a dog.’

1

This fact shows that PI relativizes virtually every constituent in a syntactic structure and is reminiscent of Adyghe, where sentential arguments are nominalized via relativization (Caponigro & Polinsky 2011). Externally headed relative clauses can be iterated showing the recursive power of the computational mechanism of human language. This is true also for LSF, as shown in (4):

(4) MAN STING DOG [ PI CHASE CAT [ PI CATCH BIRD ] ]

‘The man stings the dog who chases the cat who catch the bird.’

Properties of the relative marker. Focusing on the relative marker PI, we have distributional evidence that it is a relative pronoun rather than a complementizer. This is shown in (5a), where PI relativizes an instrument phrase and is found along the head inside the (internally headed) relative clause, and in (5b) where the head is external but PI is stranded inside the relative clause.

(5) a. IX-1 PREFER [rel LITTLE GIRL TOOTHBRUSH PI SOAK PAINT ] b. IX-1 PREFER TOOTHBRUSH [rel LITTLE GIRL PI PAINT ]

‘I prefer the toothbrush with which the little girl is painting.’

Similarly to relative pronouns in other SLs, PI shares many features with pointing pronouns rather than with wh-pronouns. PI has the extended index handshape and directional movement (plus finger aper- ture). Directionality is toward the locus of the head of the relative clause (cf. (6a)). When the directional movement of PI does not show agreement with the head, generic readings are obtained (cf. (6b)).

(6) a. IX-1 PREFER TOOTHBRUSHa PIa GIRL PAINT ‘... toothbrush the girl is painting with.’ b. IX-1 PREFER VETa PIx CURE DOG ‘I prefer vets who cure dogs.’

Analysis. Both the head and PI can be optionally found inside the relative clause (cf. (5a)). This fact paves the way for a unified analysis of internally and externally headed relative clauses in LSF, providing a strong argument in favor of a raising analysis.

Externally headed relative clauses are derived from underlying structures like the one in (7a). Distribu- tional and prosodic facts (not shown here) indicate that the head and PI occupy two distinct positions in surface structure. This can be derived either via smuggling (Collins 2005, 2006) or via two separate chains sharing the same foot (Chomsky 2008). Both techniques are designed to avoid minimality. We implement here the latter solution: PI is internally merged in spec,CP of the relative clause (cf. (7b)). The relative phrase is then re-merged within a D head, creating the external head (cf. (??)). Finally a mechanism of (partial) deletion provides the surface structure (cf. (7c)). (see Sauerland 2004 for partial deletion in relative clauses).

(7) a. [rel ... head PI ... ]

b. [CPrel head PI [ ... <head PI>... ] ]

c. [DP head [CPrel head PI [ ... <head PI>... ] ] ]

Event/situation relative clauses like (3b) are derived by merging PI in spec,CP and the entire clause into the DP (cf. (8a)). Generic readings of the kind in (6b) are derived by internal merge of the head alone into the DP, as in (8b). In these configurations agreement between the head and PI is disrupted and directionality fails to target any particular locus, nicely deriving the attested patterns.

(8) a. [DP [rel ...][CPrel PI[...<[rel ...]PI>...]]] b. [DP head [CPrel head PI [ ... <head PI>... ] ] ]

Conclusions. We investigated the morpho-syntactic properties of relative constructions in LSF showing that they are externally headed derived from internally headed structures. During the presentation we will also show that the analysis implemented here is superior to a smuggling approach (the latter does not capture the agreement pattern in (6b and 6b)).

May, 17th, 2017, Mora Maldonado

Priming and comprehension of mixed adjectives

A sentence such as The bags are heavy is thought to have both a collective and a distributive reading. On its collective reading, the sentence is used to report the total weight of bags: the plurality of bags is heavy (but each of them might not be heavy); on its distributive reading, the sentence is a report of the weight of individual bags: each bag is heavy. Note, however, that the two readings are not independent: the distributive reading entails the collective reading, whenever each bag is heavy, the plurality of bags is heavy as well. Changing the polarity of the adjective (e.g. The bags are light ) results on a reversed entailment pattern.

May, 17th, 2017, Milica Denić

Quantifier spreading

Quantifier spreading is a phenomenon of children judging a sentence such as ‘Every boy took an apple’ as false in situations where there is an apple which is not taken (nb: the sentence is false if there is a boy who didn't take an apple, not if there is an apple which is not taken by a boy). In this project, we are trying to develop a more precise understanding of the phenomenon by relating it to distributive inferences typically associated to disjunctions in the scope of universal quantifiers. We discuss how this would account for the existing child language acquisition data on the phenomenon, and lay out some predictions for future work.

May, 10th, 2017, Guillaume Dezecache (animal communication)

Affective binding in primate vocal production: studies on the early vocal behaviour of wild chimpanzees

Primate vocal production is thought to be largely inflexible, with signals bound to the expression of specific motivational and affective states. During 13 months of fieldwork with the wild chimpanzees of the Sonso community of the Budongo Forest, Uganda, I conducted a number of studies, 2 of which were dedicated to the role of affect in vocal production. In young infants of 0-to-12 months old, I looked at the extent to which two basic vocal units, the "grunt" and the "whimper-hoo" are bound to certain contexts of occurence. The data suggest that infant chimpanzees produce grunts in a variety of contexts, and that they are capable of using whimper-hoos flexibly to meet their demands, notably by altering the acoustical structure when the mother is not in immediate proximity. Altogether, these two studies point towards vocal flexibility at a very young age in wild chimpanzees.

April, 26th, 2017, Katja Liebal (animal communication)

A comparative approach to human emotion

By comparing human behavior to that of other primates, comparative psychologists aim at identifying those cognitive and communicative skills which are unique to humans, in contrast to those which we share with other species. This is a challenging enterprise, since there many conceptual, methodological and ethical constraints to explore the mental states of other species. I will argue that this is particularly true for comparative research into emotions, because of the risk to anthropomorphize great apes’ facial and bodily expressions of emotions. I will introduce different observational and non-invasive experimental methods, which we use to investigate the production and perception of emotions in different species of great and small apes, and will present some of our recent findings.

April, 19th, 2017, Sabrina Engesser (animal communication)

Vocal combinations in birds

Language’s generative capacity is one of its key characterising features. Comparative work on non-human animals investigating the evolutionary origin of combinatorial abilities has mainly focused on the vocal repertoires of singing species or primates. Whist these studies have demonstrated rudimentary combinatorial capacities outside of humans, evidence for basic phoneme-like or semantically compositional structures in non-human communication systems is rare. In this talk, I will present data on combinatorial structures in the discrete vocal repertoire of birds. Such comparative data, particularly from species distantly related to humans, can help to unveil selective drivers promoting combinatorial capacities, in addition to providing analogue examples to, and potential precursors of, language’s combinatorial layers.

April, 18th, 2017, Sam Cumming

A Taxonomy of Viewpoint Constraints

In edited film and television, the position of the camera can change between shots. During production, the camera setup is carefully monitored by continuity experts known as script supervisors, who follow rules of thumb and their own intuitions to ensure that the transitions will be seamless for the viewer. In previous work ("Conventions of Viewpoint Coherence in Film," Phil. Imprint, 2017), we argue that underlying these intuitions and rules of thumb is a system of semantic conventions of spatial coherence -- rules contributing spatial information that helps bind together the content of individual shots into a spatially connected whole, on the model of coherence relations in discourse.

Since that paper, we have made progress on mapping out the system of conventions. Specifically, we have examined spatial transitions that make reference to a prominent (usually invisible) line of action in the scene (such as the eyeline connecting a person to what they are looking at). We have discovered two broad sorts of rule. One sort dictates a specific transition of the camera between shots, while the other merely restricts the camera to a certain range of movement (and hence conveys less fine-grained spatial information). Interestingly, the specific mappings (or at least a broad class of them) involve reflecting the action line through planes formed by two axes of the camera, while the restrictions prevent the action line from passing through such planes. In the talk I will present evidence for our taxonomy and discuss reasons for empirical gaps in the space of logical possibilities.

April, 12th, 2017, Hugo Mercier

Dependency blindness and the persistence of misbeliefs

In every human culture, some misbeliefs—beliefs that are maladaptive or that have no empirical basis—have managed to spread successfully and persist across many generations. Why do people accept such misbeliefs? I suggest that the persistence of these beliefs is partly due to a flaw in the way we evaluate widely shared beliefs. When a large majority of a population shares a belief, this is usually a very good cue that the belief is accurate. One exception is when the belief was not acquired independently by the members of the majority—for instance, they all owe the belief to the same source. In some cases, people adequately discount for the presence of dependencies between the beliefs of the majority. In other cases, however, the cues to dependency are not transparent, giving rise to ‘dependency blindness.’ Many popular misbeliefs might thus persist because people think they have been acquired by the majority of the population independently, giving them solid grounds for accepting the beliefs. I will present evidence that some misbeliefs—such as religious beliefs in small scale populations—are typically presented as having been acquired independently by each member of the population, thus making them very credible.

April, 5th, 2017, Emilie Genty (animal communication)

The gestural origin of language : new insight from the study of great apes multi-modal communication

A key problem in science is to understand when and how human language evolved and in what aspects it is different from nonhuman animal communication. One way to address this question is to look for homologous traits and precursors of human linguistic abilities through the comparative study of primate communication. In terms of modality, it is unclear whether language evolved from a gestural communication system or whether it has always

been based on vocal signals. This question has been subject to a lot of debate between supporters of each theory. Theories proposing a gestural origin of language suggest that speech was preceded by a gestural phase and emphasize the similarities between ape gestural communication and human language. Indeed, the systematic investigation of great ape gestural communication has generated convergent evidence that their gestures are produced intentionally, that signalers take into account attention, comprehension and degree of knowledge of their recipient, and that key cognitive abilities necessary for the development of language were already present in the common ancestor of modern humans and great apes. While considerable progress has been made in our understanding of ape gestural communication, the fact that gestures are often produced in combination with other signals has been largely ignored. Multimodality is a core component of human language and communication, in which speech signals are routinely combined with other vocal and visual signals to convey and/or modify the speaker’s intended meaning. I will present new results from the study of bonobo multi-modal communication to illustrate how this approach could provide new insight into the study of the origin of language.

March, 22nd, 2017, Shane Steinert-Threlkeld

Pragmatic Expressivism and Non-Disjunctive Properties

Distinguishing between the concepts of compositional semantic value and assertoric content allows expressivism to be formulated as a pragmatic, not a semantic, thesis.  In this talk, I contend that the pragmatic expressivist ought to be seen as replacing the concept of assertoric content with that of a property of attitudes in their theory of communication.  Analyzing the roles played by these concepts allows for the articulation and defense of what I call the Non-Disjunctive Expressive Principle: the semantic values of sentences in context should determine a non-disjunctive property of attitudes.  After defending this Principle, I will show that the most well-developed form of pragmatic expressivism -- that of Seth Yalcin for epistemic modals -- systematically runs afoul of the principle due to the interaction of modals and disjunctions.  I will then present a new theory based on a semantics of assertability which provably satisfies the Principle and show that it accommodates other recalcitrant data concerning modals and disjunctions.  Time permitting, I will present some very preliminary experimental data concerning an unusual feature of this new theory (namely, that it derives free choice for wide scope, but not narrow scope, disjunctions).

March, 15th, 2017, Jeremy Kuhn & Carlo Geraci

Low referentiality in LSF & LIS

In this talk, we explore strategies for expressing 'low referentiality' in Italian Sign Language and French Sign Language. Cross-linguistically, there are a variety of different constructions that indicate ignorance about a discourse referent; these include epistemic indefinites (e.g. German `Maria hat ingendeinen Arzt geheiratet', `Maria has married some doctor or other') and impersonals (e.g. French `On m'a volé ma vélo,' `Someone stole my bike.'). In this talk, we explore how the patterns from LIS and LSF fit into known typologies. We discuss two existential quantifiers (SOMEONE and PERSON), two non-manual signs (frown and eye-gaze), and one syntactic strategy (the null pronoun). We show that the frown non-manual patterns nicely with epistemic indefinites in other languages, and that the null pronoun acts like an impersonal in both LIS and LSF. We discuss cases of micro-variation, where LSF differs from LIS, and also where individual signers of LSF differ from each other (but are self-consistent).

March, 15th, 2017, Nathan Klinedinst (animal communication)

Birds calls and combinatorics: methodological remarks

March, 1st, 2017, Manuel Kriz

Question Embedding, Plurals, and Homogeneity

Plural predication is known for the homogeneity property: 'Adam wrote the books' is true if Adam wrote (roughly) all of the books, but its negation is true only if he wrote (roughly) none of them. Embedded questions show in many ways analogous behaviour. For example, (1a) is intuitively true if Agatha is fully informed about who the guests were, whereas its negation (2b) conveys that she has pretty much no idea who was there.

(1) a. Agatha knows who was at the party.

     b. Agatha doesn't know who was at the party.

I argue that much of what one sees with embedded questions can be explained as direct consequence of the homogeneity of plural predication once the latter is viewed through the lense of trivalent logic.

February, 22nd, 2017, Isidora Stojanovic

Are aesthetic adjectives relative or absolute? Assessing the theoretical and empirical evidence

Gradable adjectives are customarily divided between relative (such as ‘long’ or ‘expensive’) and absolute (such as ‘full’ or ‘bent’) (McNally & Kennedy 2005, Kennedy 2007). After a brief recap of the theoretical motivations for this distinction, I present the experiments from Liao and Meskin (2016), who applied the experimental paradigm from Syrett et al. (2006) to the aesthetic adjectives 'beautiful', 'ugly' and ‘elegant’. Their findings seem to show that aesthetic adjectives do not pattern either like relative or absolute adjectives. I discuss various semantic and pragmatic factors that may be responsible for the observed results, and close with some thoughts about the directions in which inquiry should go next.

February, 8th, 2017, Mirko SANTORO

Compounds in Sign Language

In this talk I will present how compounds work in sign language. In addition to the standard type of compounds normally found in spoken languages where two roots are juxtaposed one after the other (e.g. boyfriend), I will show that sign languages allow the possibility of having simultaneous compounds, where each of the two hands articulates one of the two roots. However, the productivity of this latter option is limited by the morphological category of the two roots. Specifically, simultaneous compounds mostly involve classifier signs as one or both roots.

Among the various criteria that are normally used to identify compounds in sign language, phonological reduction is the most commonly adopted; that is, the articulation of one or both of the roots is somehow reduced with respect to the non-compound form. The rationale of the criterium is as follows: phonological reduction is a clear clue that we are dealing with a compound form; thus, those forms that entertain reduction are normally interpreted as a single lexical unit. In the second part of the talk, I provide experimental evidence that this is not the case. Specifically, sequential compounds are still treated as composed by two separate units (with some notable exceptions), while simultaneous compounds are treated as a single unit.

February, 1st, 2017, Daniël Hoek

Loose talk and conversational exculpature

In addition to conversational implicature, there is also such a thing as conversational exculpature, a pragmatic process whereby information is subtracted from, rather than added to, what the speaker literally says. This pragmatic content subtraction explains a range of linguistic phenomena. Amongst other things, it accounts for why we can say “Rob is six feet tall” without implying that Rob is between 5’11.99” and 6’0.01” tall, and why we can say “Ellen has a hat like the one Sherlock Holmes always wears” without implying Holmes exists or has a hat. For the purpose of this talk, I will focus on the application to loose talk. I presents a simple formalism for understanding conversational exculpature, specifying how, in context, the result of subtracting one piece of information from another is determined. I then show how the resulting theory can explain the central data around loose talk, including an issue involving negation that causes problems for almost every other pragmatic account.

January, 18th, 2017, Valentina Aristodemo

Temporal clauses in Italian Sign Language (LIS)

We investigate temporal constructions in Italian Sign Language (LIS) from a syntactic and semantic point of view. Syntactically, we show that these are genuine cases of adjunct subordination. Evidence for this comes from asymmetric extraction (temporal clauses adjunct islands) and prosodic features. Semantically, we show that temporal markers contain a comparative component. In this case, evidence comes from the anaphoric properties of the temporal markers, which are similar to those of comparative markers in LIS.

We claim that the surface structure of temporal constructions are derived from a relativization process followed by topicalization of the temporal clause. Following Del Prete (2008) we propose that the relevant syntactic projection hosting the temporal marker also includes a degree phrase, as in Del Prete’s analysis of before-clauses in Italian. This degree phrase allows to establish the relevant comparison between the two times. The LF structure is derived by reconstruction of the topicalized temporal clause into its original position. This structure is equivalent to that of standard comparative constructions in which the Degree Phrase that contains the subordinated clauses is adjoined to the matrix IP.

December, 21st, 2016, Mélissa Berthet (animal communication)

Alarm system of the black-fronted titi monkey

Previous work has shown that titi monkeys, Callicebus nigrifrons, emit different sequences of A- and B-calls to various predation events, and data have been consistent with the interpretation that these sequences can convey information about both the predator’s type and its location. The aim of my PhD is to investigate deeper what information are conveyed in the alarm sequences and the way receiver understand these information. We first have run predator presentations and playbacks of predators’ vocalizations to compare vocal reaction of titi monkeys to visual and acoustic cues of predator presence. Our findings are more consistent with the interpretation that titi monkey alarm call sequences encode predator type, but not location. We also have conducted acoustic analysis to investigate whether the A and B-call can be further discriminated into acoustic variants and whether they are given in context-specific ways. Our results suggest that A-calls are relatively context-specific by exclusively referring to dangerous, non-terrestrial predators. B-calls, in contrast, are given to wider range of events, but these calls can be subdivided into acoustic variants. Finally we run playback to investigate how titi monkeys understand alarm sequences of conspecifics, but results are not available at the moment.

December, 14th, 2016, Cornelia Ebert

The semantic behaviour of co-speech gestures and their role in demonstrative reference

In my talk, I will relate the study of speech-accompanying gestures (e.g. McNeill 1992, Kendon 2004) to discussions about ‘multidimensional meanings’ (e.g. Potts 2005) and argue that by default, gesture meaning enters into composition as supposition-like non-at-issue material. I will compare my predictions to those of Schlenker (2016), who argues that gestures are a special kind of presuppositions, i.e. cosuppositions. I focus on iconic and pointing gestures.

I will furthermore show that demonstratives can be analysed as semantically vacuous entities that only function as ‘dimension shifters’ from the non-at-issue to the at-issue dimension. The („lexical“) semantics of a gesture is taken to be a rigid concept such that not the demonstrative expression, but the gesture itself is responsible for the observed rigid interpretation. This treatment of a gesture (and the corresponding temporally aligned demonstrative) then derives Kaplan’s analysis of demonstratives + pointing gesture.

The formal treatment of (pointing) gestures and the meaning contribution of their interplay with speech finally also accounts for Kripke’s distinction between speaker’s referent and semantic referent and has the potential to further our understanding of the attributive/referential distinction brought up by Keith Donellan (1966).

December, 7th, 2016, Camille Coye (animal communication)

Vocal combinations in guenon communication

Duality of patterning is a core feature of language which relies on the combination of sounds at two levels, phonology and morphosyntax. This capacity is central to our capacity to generate an infinity of meanings using a limited number of elements. Sound combination occurs in the daily communication of animals from various taxa and, for several decades, ethologists and psychologists have been studying capacities for vocal combination in animals using a comparative approach with language. However evidence for non-random patterns of combination leading to consistent changes in utterances’ meaning often lack. Recently, observational studies suggested the presence of meaningful and context-dependent sound combination in two sympatric species of African forest guenons, Campbell’s monkeys (Cercopithecus campbelli) and Diana monkeys (C. diana). In particular, male Campbell’s monkeys give one alarm call type (‘Krak’) to leopards and seem to combine it with an “oo” unit (i.e. ‘Krak-oo’) to signal unspecific danger. Similarly, female Diana monkeys give four social calls emitted alone or merged into six combinations. In combinations, the initial call conveys information about social context, while the final call conveys information about caller’s identity. In order to assess the combinatorial nature of these calls (i.e. whether they result from the merging of distinct units) and their relevance to receivers, we performed playback experiments using natural and artificially combined calls for both male Campbell’s and female Diana monkey calls. Responses indicated that monkeys attended to the way calls were combined regardless of contextual origin and confirmed the presence of meaningful call combination in our study species. The results obtained will be discussed, together with other recent findings on vocal combination in animals, to propose a methodological and theoretical reflection on future steps required to further explore the nature of animal combinatorics.

November, 30th, 2016, Jana Hosemann

Agreement in sign languages and spoken languages: same or different? – Evidence from a neuro-linguistic perspective

The phenomenon that is called “agreement” in sign languages is very different from verb agreement in spoken languages. At least, when viewed from the surface. Sign language agreement is – in principle – a location overlap between the beginning/ending of the verbs path movement and the locations associated with the verbs arguments (i.e. subject and object). In other words, if the subject is associated with the area on the right side of the signer (indicated by 3a), and the object is associated with the area on the left side of the signer (indicated by 3b), the verb moves from right to left. Compare (1):

(1) TOM IX3a SARAH IX3b 3aHELP3b

    ‘Tom helps Sarah.’

In this talk, I will show evidence that supports the idea that – despite the obvious differences in the realization of agreement in sign languages and agreement in spoken languages – both phenomena are cognitively processed similarly. And that therefore, agreement is a modality-independent morpho-syntactic phenomenon. I will also provide neuro-linguistic evidence that argues against this position. Thereby, I hope to encourage a discussion on the linguistic status of sign language agreement.

November, 30th, 2016, Cathy Crockford (animal communication)

How much does cognition influence chimpanzee single and combined vocalisations?

In stark contrast to human speech, animal signals are thought to be tied to emotional states and hence tied to fixed social and ecological contexts that precipitate certain emotions. This leaves 1) little flexibility in signal production within contexts and 2) little cognitive influence on signal production. We argue that these assumptions are premature. First, prelimary data comparing the usage of call combinations across different subspecies of chimpanzees, from the Tai Forest, Ivory Coast and the Budongo Forest, Uganda, indicates usage difference. Call combinations in a particular context, within and between group aggression, are typically composed of different call types. Also, in another context, greeting contexts, the order of call cominations is reversed between subspecies.  Both examples indicate changes in call combination usage across populations.

Second, we address the influence of cognition on signal production. Animal signallers are assumed not to possess the capacity to take others’ perspectives or knowledge into account when signalling, a crucial ability for the evolution of language. We argue this assumption is premature. Recent studies show primate alarm signalling to threats changes according receivers’ risk. However, for these and other studies examining mind-reading, the confound that signallers read receivers’ behaviour rather than their mind is hard to exclude. Here, we excluded behaviour reading by simulating a receiver, using a playback of a conspecific’s call. We tested if wild chimpanzees’ signalling to a snake model depended on the simulated receivers’ knowledge of the snake, whether simulated receivers emitted snake-related calls (indicating knowledge), or acoustically similar non-snake related calls (indicated ignorance). Signallers showed greater vocal and non-vocal signalling effort and speaker-directed monitoring when simulated receivers had not emitted snake-related calls. Informing and monitoring effort was greatest when simulated receivers apparently lacked information about the snake model and was not explained by arousal or habituation to the snake model , or the influence of the simulated receivers’ identity or behaviour. The results go further other studies, showing that chimpanzees are motivated to fill another’s information gap when facing a threat. For the origins of language, we suggest the language faculty evolved from a pre-existing hominoid socio-cognitive capacity, before the advent of more language-specific features, such as recursive syntactic structure.

November, 23rd, 2016, Gary Lupyan

From perception to symbolic thought: how language augments human computation

A common assumption in psychology and linguistics is that words map onto pre-existing meanings. I will argue that this position is mistaken and that words play a much more central role in creating meaning than is generally acknowledged. Using behavioral, EEG, and patient data, I will present evidence that even in linguistically savvy adults, the use of language actively modulates “nonverbal” mental processing starting with low-level perception. On the presented view, many of the unique aspects of human cognition stem from the power of words to flexibly create categories from perceptual representations, allowing language to act as a high-level control system for the mind. This view has immediate consequences for understanding the cognitive consequences of learning and using natural language and for questions concerning linguistic relativity.

November, 16th, 2016, Milica Denic

Experimental studies of intervention effects in NPI licensing

It is generally accepted that the relevant semantic requirement for the NPI licensing is downward-entailingness (DE-ness). The ungrammaticality of the sentence in (1) is thus unexpected, as the NPI 'any' is at a DE position there.

(1) *Mary didn’t give every lion any chocolate.

This phenomenon is known as intervention effects in NPI licensing. An influential theoretical approach to this phenomenon (Chierchia 2004, 2013) proposes that it is indirect scalar implicatures which cause intervention effects by disrupting the DE-ness of the environment.

In this talk, we will discuss the experimental data we obtained attempting to evaluate this proposal.

November, 9th, 2016, Amir Anvari and Mora Maldonado

Redundancy and disjunctions in online processing

The unacceptability of sentences such as (1) has been used to motive a theory of acceptability based on a ban against redundant material.

(1) #Unless Mary is pregnant, she is not pregnant and John is happy

(2) a. Either Mary is pregnant or she is not and John is happy

    b. Either Mary is pregnant or John is happy

This generalization has been threatened by the acceptability of disjunctive sentences such as (2a), which seem to be equivalent to the simpler (2b). If redundant material is generally prohibited, why both disjunctive sentences are acceptable? Mayr and Romoli (2013) offer a solution to this issue according to which after exhaustifying (2a), the sentence is indeed not redundant.

As is assumed by the theory, to avoid redundancy, people should be always forced to enrich the meaning of disjunction in sentences like (2a), interpreting the disjunction with an exclusive meaning. During comprehension, this strengthening could operate either at the point at which or is first encountered, or subsequently, as a consequence of a reanalysis. The acceptability of sentences such as (2a) might therefore involve two different underlying strategies.   

In this talk, we will show experimental data suggesting that, contrary to what is predicted by the theory, strengthening does not seem to help to avoid redundancy. Instead, participants who are biased towards an exclusive interpretation of disjunction are more sensitive to redundancy than participants who display an overall weak interpretation of disjunction.

November, 2nd, 2016, Jeremy Kuhn

A new synonymy problem for E-type theories

In E-type theories of anaphora (Evans 1980, Elbourne 2005), cross-sentential pronouns and donkey pronouns are analyzed as definite descriptions, so the Logical Form of the pronoun 'it' in (1a) is the definite description 'the donkey'. This analysis captures the equivalence in meaning of sentences with pronouns and sentences with definite descriptions, as in (1).

(1)a. Christopher owns a donkey. It brays.

b. Christopher owns a donkey. The donkey brays.

In this talk, I present an argument against E-type theories, using new data from sentences involving 'same' and 'different'. In a nutshell: the first conjuncts of (2a) and (2b) are true under exactly the same conditions, yet they differ in what discourse referents they make available. The ungrammaticality of the pronoun in (2b) is unexpected on an E-type theory, since a full definite description can reside happily in the same context, as in (3b).

(2)a.   All the boys read a different book, and all of them liked it.

b. * No boys read the same book, {and/but} all of them liked it.

(3)a. All the boys read a different book, and all of them liked the book they read.

b. No boys read the same book, but all of them liked the book they read.

In this talk, I develop this argument within two specific E-type analyses: Heim (1990) and Elbourne (2005). After demonstrating that neither framework can capture these facts, I offer some perspectives on treatment of these sentences within dynamic semantics.

October, 26th, 2016, Connie de Vos

Linguistic cues enabling rapid conversational turn-taking in sign

Face-to-face interaction is the core ecological niche for language: interaction is the point of origin for language emergence, acquisition, and evolution. To understand the linguistic capabilities of our species it is necessary to analyse language in interactive contexts. Across the world’s languages, interlocutors in spontaneous conversation time their utterances to minimise gaps and overlaps between consecutive turns, resulting in remarkably fast turn-transitions (mode = 200ms; Stivers et al. 2009). Recent analyses indicate that the same turn-timing pattern extends to signed languages, provided that signers consider the end of the turn-final stroke (i.e. the end of the lexically-specified movement of signs) as the end of the turn itself (de Vos et al. 2015). Following this, addressees in signed conversation must anticipate current signers’ upcoming turn-ends to plan and initiate their responses with minimal-gap-minimal-overlap timing (similar to spoken conversation; Sacks et al. 1974; De Ruiter, Mitterer & Enfield 2006). I will present experimental results indicating that signers of Sign Language of the Netherlands predict the end of turn-final strokes in NGT with accuracy comparable to Dutch speakers’ predictions about turn-ends in Dutch (Magyari & de Ruiter 2012), and discuss the role of various linguistic cues driving signers’ early responses for questions in particular.

De Ruiter, J. P., Mitterer, H., & Enfield, N. J. (2006). Projecting the end of a speaker's turn: A cognitive cornerstone of conversation. Language, 515-535.

Stivers, T., Enfield, N. J., Brown, P., Englert, C., Hayashi, M., Heinemann, T., ... & Levinson, S. C. (2009). Universals and cultural variation in turn-taking in conversation. Proceedings of the National Academy of Sciences, 106(26), 10587-10592.

Magyari, L., & de Ruiter, J. P. (2012). Prediction of turn-ends based on anticipation of upcoming words. Frontiers in psychology, 3, 376.

De Vos, C., Torreira, F., & Levinson, S. C. (2015). Turn-timing in signed conversations: Coordinating stroke-to-stroke turn boundaries. Frontiers in Psychology, 6: 268. doi:10.3389/fpsyg.2015.00268.

October, 5th, 2016, Daniel Rothschild

Placing the Problem of Permission

Lewis (1979) described the 'problem of permission' as the the problem of making sense of how to update the sphere of permissible worlds in light of new permissions. Permitting an action allows into the sphere of permissible worlds some but not all worlds in which that action is performed. I will discuss some new attempts at approaching this problem using the framework of truthmaker semantics (as developed recently by Kit Fine and Stephen Yablo). I will also discuss the relationship between this problem of permission and the semantics of the possibility modals (such as 'may' and 'can') that can be used to express permission.

September, 28th, 2016, Amir Anvari

ANCHORED INTERPRETATION: on the formal pragmatics of co-nominal pointing

In this talk we report ongoing research on the interpretive consequences of those occurrences of pointing gestures that are temporally aligned with quantificational noun phrases, or co-nominal pointing for short. For instance, if an assertion of (1)a is accompanied by a pointing gesture toward an individual, say John, the inference is triggered that John is one of the relevant students and he smokes. On the other hand, if the same gesture accompanies (1)b, the inference is triggered that John is one of the relevant students but he does not smoke.

(1)a. At least one student smokes.

b. Not every student smokes.

It will be argued that co-nominal pointing comes with certain use-conditions as well as a rich repertoire of possible inferences that in general depend on (i) the number of individuals pointed to, (ii) the logical properties of the relevant QNP, and (iii) the logical syntax of the sentence in which the relevant QNP is embedded. Theoretically, three hypotheses pertaining to the mechanics of co-nominal pointing will be explored and evaluated.

The domain restriction hypothesis builds on the mechanism of quantifier domain restriction (Stanley & Szabó 2000) and analyses co-nominal pointing as triggering the presupposition that the meaning of the sentence S containing the relevant QNP with restriction to the individual(s) pointed to is equivalent to the meaning of S without such restriction. It follows that in evaluating the sentence, the hearer can safely ignore those individuals not pointed to, as there will be no interpretive repercussions relative to the common ground.

According to the exemplification hypothesis (in its simplest version) co-nominal pointing provides an example of a certain set of individuals involved in the interpretation of the relevant determiner. Intuitively, in (1)a the pointing gesture toward John is analysed as triggering the inference that “John is one of them”, where the pronoun “them” gets is value (i.e., the set of students who smoke) from (1)a by the same mechanism that is responsible for plural discourse anaphora in (2), e.g. along the lines of Nouwen (2003).

(2)Some students in my class are lazy. They [= the students in my class that are lazy] will fail the exam.

Finally, we sketch a hypothesis based on recent work on truth-maker semantics (Yablo 2014, Fine forthcoming). One version of this hypothesis claims that by pointing the speaker signals that the individual(s) pointed to is a “witness” to the sentence in some sense, and according to the second version by pointing the speaker signals that his utterance is in some sense “about” the individual(s) pointed to.

We close by putting this line of inquiry in the context of pointing in sign language semantics, asking in particular whether co-nominal pointing reveals a mechanism that is grammaticalised in sign language or whether the function of co-nominal pointing in speech can be replicated in sign language by, e.g., head tilt or eye gaze.

July, 13th, 2016, Salvador Mascarenhas

The conjunction fallacy as reasoning about ignorance

In a seminal paper from 1983, Tversky and Kahneman showed that, with certain choices of content, human reasoners will consider a statement of the shape `A and B' to be more probable than `A'.  According to probability theory, this is a fallacy, for the probability of a conjunction of independent propositions has to be less than or equal to the individual probabilities of its conjuncts.  Tversky and Kahneman proposed an account in terms of representativeness: instead of asking themselves which statement is more probable, reasoners ask themselves which property instantiated in each of the two statements is the topical individual more representative of / a more typical example of.  When the setup and content of the task make the property in the conjunctive statement a more typical property for the topical individual, reasoners will pick the conjunctive statement, providing a reasonable answer to a question about typicality, but an irrational answer to a question about probabilities.

In this talk I argue that the conjunction ``fallacy'' can alternatively be understood as (mostly) rational thought acting on non-obviously pragmatically strengthened premises.  I present arguments in favor of this approach building on independent theories of epistemic (ignorance) implicatures.  I further show how a fully explicit version of this account can be derived from Kratzer's (1991) theory of modality (building on Lewis, 1981), a bare-bones theory of epistemic implicatures, and a small collection of plausible epistemic biases.  In a nutshell, I argue that the `A' option is in fact interpreted as `A and possibly B and possibly not B'.  I show that, under two weak assumptions about the modal base, if we assume that `B' is considered a better possibility (in Kratzer's sense) than `not B', then `A and B' will also be a better possibility than `A and possibly not B'.  I conclude with an independent argument against Tversky and Kahneman's representativeness account, arguing that not only do there exist promising alternative explanations of the phenomenon in terms of pragmatics and modality, the representativeness account itself struggles with examples of the conjunction fallacy that are unproblematic for a pragmatic approach.  These are examples of conjunction fallacy setups where there isn't a clear topical individual with respect to which representativeness/typicality questions can be asked.

July, 6th, 2016, Brian Buccola

A surface-scope analysis of authoritative readings of modified numerals (joint work with Andreas Haida)

This talk concerns a well-known puzzle regarding the interpretation of superlative modified numerals ("at least n", "at most n") in the scope of an existential modal ("can", "allow"). The puzzle is this. A sentence like (1) has a so-called "authoritative reading", characterized by two kinds of inferences: (i) an upper-bound (UB) inference, viz. that you're not allowed to draw more than 3 cards, and (ii) a free-choice (FC) inference, viz. that you're allowed to draw any number of cards in the range [0,3] (Geurts and Nouwen 2007; Büring 2008; Nouwen 2010; Penka 2014; Kennedy 2015). On standard assumptions about the meanings of "allow" and "at most 3", however, a surface-scope analysis predicts only a weak literal meaning, from which neither a UB nor an FC inference logically follows.

(1) You're allowed to draw at most 3 cards.

What's more, "at least n" does not give rise to an analogous authoritative reading: (2) cannot be used to convey that you're allowed to draw 3 or more cards, with FC in the range [3, ...], and a lower bound that prohibits drawing 2 cards or fewer.

(2) You're allowed to draw at least 3 cards.

The puzzle, then, has two parts: (i) why does (1) have an upper-bounded authoritative reading, and (ii) why does (2) not have an analogous lower-bounded authoritative reading?

I will discuss a decompositional, inverse-scope-based solution, due to Penka (2014), and present some novel data that I argue are problematic for such an account. I will then offer a new (but partial), surface-scope solution based on a conservative amendment of the recursive exhaustification approach to FC disjunction as laid out in Fox 2007. Time permitting, I will discuss how this proposal relates to recent work devoted to deriving the ignorance inferences associated with superlative modified numerals (in particular Schwarz 2016).

June, 29th, 2016, Jakub Szymanik

Towards mental logic

Natural language contains an abundance of reasoning patterns. Historically, there have been many attempts to capture their rational usage in normative systems of logical rules. However, empirical studies have repeatedly shown that human inference differs from what is characterized by logical validity. In order to better characterize the patterns of human reasoning, psychologists have proposed a number of theories of reasoning. In this paper, we combine logical and psychological perspectives on human reasoning. We develop a framework integrating Natural Logic and Mental Logic traditions. We model inference as a stochastic process where the reasoner arrives at a conclusion following a sequence of applications of inference steps (both logical rules and heuristic guesses). We estimate our model (i.e. assign weights to all possible inference rules) on a dataset of human syllogistic inference while treating the derivations as latent variables in our model. The computational model is accurate in predicting human conclusions on unseen test data (95% correct predictions) and outperforms other previous theories. We further discuss the psychological plausibility of the model and the possibilities of extending the model to cover larger fragments of natural language.

June, 22nd, 2016, Fabio del Prete

Complex events in Sicilian

Sicilian exhibits a complex verbal construction, called Doubly Inflected Construction (DIC; Cruschina 2013), as in (1):

(1)  Vaju        a         accattu    a    cicoria.

      go[1SG] PREP buy[1SG] the chicory

      ‘I go buy chicory.’

In DIC, a motion verb V1 from a class including "vèniri" ‘come’, "passari" ‘pass’, and the causative verb "mannari" ‘send’ is followed by an eventive verb V2, with V1 and V2 being inflected for the same person and tense features (Feature Matching). DIC differs from the Hypotactic Construction (HC) in (2), in which V1 is inflected and V2 occurs in an infinitival clause complement of V1:

(2)  Vaju        a         accattari  a    cicoria.

      go[1SG] PREP buy[INF]  the chicory

      ‘I go to buy chicory.’

Based on syntactic and semantic evidence, it has been argued that DIC, unlike HC, is a monoclausal construction in which V1 and V2 form a complex predicate referring to a single event. The monoclausal view is argued for by Cardinaletti and Giusti (2001), who further claim that in DIC only V2 has full lexical content and interpreted inflectional features, while V1 is partially desemanticized and inherits its features from V2.

In this talk I consider previously neglected DICs with "mannari" which are problematic for Cardinaletti and Giusti's analysis, as in (3):

(3)  Ti          mannu       a        ddicu        na    cosa.

     To-you  send[1SG] PREP say[1SG] DET thing

     ‘I send someone say something to you.’

Here the person feature on "mannu" is the same as the one on "ddicu", i.e., first singular (by Feature Matching). However, this feature is interpreted only on "mannu", since the agent of "ddicu" is not the speaker but some other person whom the speaker sends to say something to the hearer. Based on this and other observations (e.g., stative verbs are generally excluded from V2), I argue for a view of DIC as an asymmetric serial verb construction (in Aikhenvald's 2006 sense), where V1 is a lexical verb contributing a motion event and projecting a theme and a goal argument, and argument-sharing occurs between V1 and V2. On the analysis I propose, only V1 has interpreted features, while the features on V2 are triggered by a morphosyntactic mechanism requiring identity between the two feature-sets and they remain invisible to semantic interpretation. The semantic analysis is based on an operation of event concatenation and I discuss two possible implementations of it in a compositional framework: an extension of Landman's (2008) approach to verb clusters in Dutch and German, which makes use of function composition, and an application of Champollion's (2015) event framework, in which function application is assumed as the only operation of meaning composition.

June, 15th, 2016, Vincent Janik (animal communication)

Complexity and Meaning in Marine Mammal Communication

One of the most significant steps in the evolution of complexity in acoustic communication was the acquisition of vocal learning skills. Vocal learning in the sense of copying novel sounds is rare in mammals and only occurs in humans, marine mammals, elephants and bats. The seemingly most convincing evidence for vocal learning comes from animals copying human speech. However, this can be misleading if calls in their natural repertoire resemble human utterances. We have developed a training method that allows testing vocal learning skills more objectively by using existing sounds in the animal’s repertoire as a starting point before progressing to artificial stimuli. Using this method, we could demonstrate vocal learning of formants in grey seals, and quantify vocal learning abilities in bottlenose dolphins. While the necessity for vocal learning in grey seals is a little unclear, bottlenose dolphins use these skills in the development and copying of signature whistles. These whistles are individual identification calls and are used when animals meet at sea or are trying to stay together. Each animal develops its own unique signature whistle early in life. The fundamental frequency modulation pattern of this whistle does not correlate with size, age, sex or relatedness, making it a relatively arbitrary signal. Signature whistles are produced in a specific temporal delivery pattern that makes them recognisable as signatures. Other animals occasionally copy signature whistles and can use this copying to address the whistle owner. Copying events in interactions occur in a specific time window after the owner produced a whistle, but dolphins also copy signature whistles of absent animals, most likely in an attempt to make contact. Signature whistles are potentially referential when copied by others, but further tests are necessary to fully explore the use of reference in marine mammal communication. Marine mammals show a remarkable range of cognitive skills affecting their communication, but most data come from artificial tests in captive settings. The exploration of how animals use these skills in their own communication system is still in its infancy but revealed patterns that highlight the role of vocal complexity in social interactions.

June, 8th, 2016, Jean-Remy Hochmann

Same but not Different?

The concepts same and different have been the topic of many studies as the paradigmatic example of abstract relations, which are necessary to the development of language and logical reasoning. They are also interesting for another reason: they are linked by negation: same is not different, and different is not same. Do infants represent these two relations? Do they represent the logical link between them? In a series of anticipatory looking and pupillometry experiments, I investigate infants’ ability to represent these abstract relations. In 4- to 14-month-olds, we find strong evidence in favor of a representation of the relation same. Evidence for a representation of different are much weaker. I will suggest that this pattern of success and failure tells us something about the limitations of infants’ language of thought and the format of their representations.

May, 18th, 2016, Natasha Abner

Distinguished categories: Structured Iconicity in Established and Emergent Sign Systems

A fundamental distinction in human language is that between labels for actions and relations between objects (verbs) and labels for the objects themselves (nouns). Beginning with the pioneering work of Supalla and Newport (1978) on American Sign Language (ASL), the distinction between nouns and verbs in a number of sign languages has been observed to be morpho-phonologically transparent (Johnston 2001, Hendriks 2008, Hunger 2006, Kimmelman 2009, Kubus 2008, Pizzuto and Corraza 1996, Schreurs 2006, Voghel 2005): language-specific phonological parameters are reliably found with one, but not the other, of these grammatical categories. This talk uses the noun-verb alternation in an established sign language (ASL) and its development in the emergent Nicaraguan Sign Language (NSL) as a way to understand another pervasive property of sign languages, iconicity. Iconicity is the motivated, non-arbitrary connection between language form and language meaning. Once considered to be a challenge to the linguistic status of sign languages, the case study of the noun-verb alternation shows that iconic properties are fully incorporated within and manipulated by the linguistic system.

May, 4th, 2016, Napoleon Katsos

The acquisition of quantifiers across languages and populations: empirical generalisations in search for linguistic and cognitive explanations

Learners of most languages are faced with the task of acquiring words to talk about number and quantity. Much is known about the order of acquisition of number words and the cognitive and perceptual systems and cultural practices that shape it. Substantially less is known about the acquisition of quantifiers. In this presentation, I will consider systems and practices that support number word acquisition in order to determine that their relevance to quantifiers is limited. Instead, I will propose that a major constraint in the acquisition of quantifiers comes from their rich and varied meaning. As part of COST Action A33, we investigated competence with the expressions for ‘all’, ‘none’, ‘some’, ‘some…not’ and ‘most’ in 31 languages, representing 11 language types, by testing 768 5-year-old children and 536 adults. The findings revealed four dimensions of the meaning and use of quantifiers that constrain the order of acquisition in similar ways across languages in our sample. The presentation will conclude with a discussion about the linguistic and cognitive underpinnings of these empirical generalisations, informed by data from the acquisition of the same quantifiers by Spanish-speaking children with SLI and/or ASD.

April, 27th, 2016, Rory Tornbull

Intonation

In many of the world's languages, information structure can be conveyed through the use of prosody and intonation. In this talk, I present research examining the prosodic marking of focus in four typologically unrelated languages: American English, Paraguayan Guaraní, Moroccan Arabic, and K'iche'. The same interactive task was used in all four languages, in their respective countries, to gather spontaneous utterances. Crucially, the context of each utterance was varied systematically to elicit different focus conditions and differing degrees of contextual predictability. The data were phonologically annotated and phonetically analyzed. The results show that the focus-marking strategies deployed by each language are different, suggesting that the prosody-semantics interface to some degree language-specific. Further, the differences between the languages did not correlate with other aspects of the languages' prosodic typology (such as head- versus edge-marking), suggesting that focus marking is orthogonal to aspects of prosodic typology. These results demonstrate the value of cross-linguistic comparisons in theory-building for both sound and meaning.

March, 23rd, 2016, Alexandra Simonenko

Agreement syncretisation and the loss of null subjects in Medieval French: quantitative and structural models

In this paper we examine quantitative and structural relations between the richness of verbal subject agreement and the availability of null subjects, known as Taraldsen’s Generalisation (Taraldsen 1980, Rizzi 1986, Adams 1987), from the point of view of grammar change in Medieval French based on corpus data. The original generalisation is stated as a categorical implication, namely, that a language having rich (i.e. non-syncretic) subject agreement implies the possibility of non-expression of subjects. Synchronic structural accounts of null subjects commonly involve as a key ingredient a functional head hosting the verb and specified with person features, which, depending on the approach, either enters into a feature identification relations with a pro or, if no pro is postulated, itself serves to fill the verb’s subject argument slot (see Sheehan (to appear) for an overview). Semantically, null subjects have been conceived of as minimal pronouns, in the sense of Kratzer (2009), which need to acquire φ-features from the verbal head which binds them (e.g. Alonso-Ovalle and D’Introno 2000). In this optics, a diachronic loss of rich agreement and null subjects can be naturally modelled in Kroch’s (1989) fashion as a competition between an old grammar with a verbal head specified for person features and a new grammar without such head. In particular, for Medieval French agreement syncretisation has been considered to be a trigger for the null subject loss by Ewert (1943), Vennemann (1975), Mathieu (2006). However, Roberts (1993, 2014) rejects a direct connection between the two phenomena because of an apparent temporal lag: he dates agreement syncretisation in Medieval by the XII c., based on an extrapolation from written to oral data in Foulet (1935/36), and the completion of the loss of null subjects by the XVI c. In contrast, a number of more recent phonological reconstructions suggest that phonological restructuring of the verbal agreement was underway throughout Medieval French (Buridant 2000, Morin 2001, De Jong 2006). It therefore seems premature to abandon the hypothesis about an immediate grammatical relation between non-syncretic agreement and null subjects.

As empirical contribution, this paper presents the first, to our knowledge, quantitative measure of agreement syncretisation in Medieval French. We also show that agreement syncretisation and the disappearance of null subjects proceeded at the same rate. On the Constant Rate Hypothesis of Kroch (1989), which states that a given grammatical change has the same rate in different contexts, our results support approaches that view non-syncretic agreement and null subjects as two manifestations of the same grammar. We associate such grammar with the presence of a verbal functional head specified with person features which identifies a pro. Finally, we propose two alternative models of the propagation of the competing grammar – Production and Perception-based – and show that the corpus data support the former (contra Vance and Sprouse 1999). Our results based on diachronic data thus bear on the general theory of null subject syntax and acquisition.

March, 23rd, 2016, Pawel Fedurek (animal communication)

The complex acoustic structure of a chimpanzee pant hoot reflects its complex social function

Pant hoots are species-specific long-distance calls commonly produced by chimpanzees, especially males. The call has a distinct acoustic structure consisting of four phases including both low- and high-frequency elements. My research has aimed to identify the social functions of pant hooting as well as to explore the link between the complex acoustic structure of this call and the versatile roles it plays. Regarding the function, my research showed that these calls play several important social roles, including regulating grouping dynamics in an unstable fission-fusion chimpanzee society, signaling caller’s dominance status and social bonds between males. Regarding the acoustic structure, the four phases of a pant hoot correlate differently with caller’s identity, age, dominance rank, and the context of call emission, and this information is directed at both nearby and distant receivers. The call, with the two low-frequency initial phases leading to the high-frequency climax, is also shaped in a way that facilitates chorusing - an important bonding behavior among males. Pant hooting also requires the caller to immediately switch between phases with extreme frequencies, a behavior that may reflect male’s physical condition. Indeed, my data show that males with high testosterone levels produce pant hoots with high-frequency climaxes. I conclude that the complex acoustic structure of a pant hoot reflects the complex social function it fulfills.

March, 16th, 2016, Jacopo Romoli

A challenge to parsing-based approaches to presuppositions and triviality

In this talk, I use antecedent-final conditionals to formulate two problems for parsing-based theories of presupposition projection and triviality of the kind given in Schlenker’s (2009). I’ll show that in these cases parsing-based theories predict filtering of presuppositions where there is in fact projection, and triviality judgments for sentences which are in fact felicitous. More concretely, these theories predict that presuppositions triggered in the antecedent of antecedent-final conditionals will be filtered (i.e. will not project) if the negation of the consequent entails the presupposition. But this appears wrong: (1) intuitively presupposes that John is in France, contrary to this prediction.

(1) John isn’t in Paris, if he regrets being in France.

Likewise, parsing based approaches to triviality predict that material entailed by the negation of the consequent will be redundant in the antecedent of the consequent; but (2) is intuitively felicitous, contrary to these predictions.

(2) John isn’t in Paris, if he’s in France and Mary is with him.

Importantly, given that the trigger appears in sentence-final position, both incremental left-to-right and symmetric versions of such theories make the same predictions.

These data constitute a challenge to the idea that presupposition projection and triviality should be computed on the basis of parsing. This issue is important because it relates to the more general question as to whether presupposition and triviality calculation should be thought of as a pragmatic post-compositional phenomenon or as part of compositional semantics, as in the more traditional dynamic approaches. I discuss two kinds of possible solutions. The first allows us to maintain the parsing based pragmatic approach; it is based on an analysis of conditionals which incorporates a presupposition that their antecedent is compatible with the context, together with a modification to Schlenker’s (2009) algorithm for calculating local contexts so that it takes into account presupposed material. As I will discuss, this solution works within a framework broadly similar to that of Schlenker’s (2009), but it doesn’t extend in an obvious way to other parsing-based accounts like (e.g., Fox 2008, 2012, Chemla & Schlenker 2012, Chemla 2010). The second solution abandons parsing altogether for hierarchical order based approaches to presuppositions and triviality, as in traditional dynamic approaches, as well as more recent dynamic approaches (e.g. Chierchia 2009, Rothschild 2008, 2011), and trivalent approaches based on hierarchical order (as discussed in George 2008 and Romoli 2012).

March, 9th, 2016, Jeremy Kuhn

Telicity and iconic scales in ASL

In a series of papers (Wilbur 2003, i.a.), Ronnie Wilbur shows that a number of sign languages display a non-arbitrary form-to-meaning correspondence in the verbal lexicon: telic verbs end with sharp deceleration (‘end-marking’); atelic verbs do not. In ASL, Wilbur also shows that the phonetic form of a verb may be manipulated with semantic effect. In this talk, I provide an analysis of these facts in terms of structural iconicity, where the interpretation of a sign preserves abstract structure of the form of the sign. I follow Kennedy and Levin 2008 in assuming that the meanings of change-of-state verbs are derived from scales; I argue that verbs in ASL iconically represent these scales, and that end-marking on telic verbs is the iconic representation of the maximum of a closed scale.

February, 24th, 2016, Heather Burnett

Signalling Games, Sociolinguistic Variation and the Construction of Style

This presentation introduces social meaning games (SMGs), which are developed for the analysis of the strategic aspect of sociolinguistic variation (in the sense of Labov 1963, Labov 1966, Weinreich et al. 1968, et seq.). While remarks have been made (eg. Clark 2011:88) about the potential usefulness of game theory in the analysis of the meaning of variable linguistic phenomena (for example, variable use of the English (ING) suffix (1)), a general framework uniting variationist sociolinguistics with game theoretic pragmatics has yet to be developed.

(1)I’m workin’ on it vsI’m working on it.

I propose that such a unification is possible through the integration of the third wave approach to the meaning of sociolinguistic variation (see Eckert 2012) with signalling games (Lewis 1969), which are commonly used in the game-theoretic analysis of Gricean reasoning (Benz et al. 2005, Franke 2009, Clark 2011, Frank & Goodman 2012, among many others). I define the games and then show certain results concerning the framework’s predictions for contextually-based style-shifting, as exemplified by three empirical studies:

1. Labov (1966)’s study of the social stratification of (ING, i.e -in’ vs -ing) in New York City.

2. Labov (2012)’s study of President Obama and Sara Palin’s use of (ING) in formal vs informal settings.

3. Gratton (2015)’s study of the use of (ING) by non-binary individuals (i.e., individuals who do not identify as either male or female) in their home vs a public coffee shop.

Based on these examples, I conclude that SMGs constitute useful tools for the analysis of the meaning of sociolinguistic variation and, more generally, for the integration of quantitative sociolinguistics into the broader domain of formal pragmatics.

February, 17th, 2016, Doreen Georgi

On the possessor case alternation in Udmurt

In this paper we investigate the case split on the possessor in Udmurt. Traditionally, the choice between ablative and genitive possessor case is said to be driven by the grammatical function (GF) of the XP containing the possessor. We argue that the case split is not driven by GFs; rather, it is determined by the case value of the XP that contains the possessor.

Importantly, there is no evidence which points to a possessor raising analysis in Udmurt. Instead, we present an analysis according to which the possessor always bears genitive but may be assigned another structural case by an external head. Due to a morphological constraint, stacked case features fuse into a single feature set in the postsyntactic morphological component. If accusative and genitive stack on the possessor, only the default semantic case marker, i.e., the ablative marker, can realize the resulting feature set. In any other context the genitive marker is chosen. We thus claim that there is no abstract ablative on the possessor; instead, the morphological ablative marker realizes a combination of two abstract structural cases.

February, 10th, 2016, Gladez Shorland (animal communication)

Social learning and inference from social knowledge in bonobos (Pan paniscus)

As part of a broad project investigating the evolutionary origins of human language, we adopt a comparative approach, looking into specific aspects of communication in closely related primates in order to better estimate when certain cognitive capacities underlying communication appeared in the lineage. The aim of this research is to investigate the inferential abilities of bonobos, more specifically, whether or not they are able to draw inferences from previously acquired social knowledge upon hearing vocalisations. In a captive group of bonobos at la Vallée des Singes primate park in France, we investigate whether bonobos can infer where food is, based on calls from individuals with particular food preferences. This question is based on the assumption that bonobos can acquire and memorise information pertaining to different food preferences of group members. The first step in this study was therefore to investigate whether or not this is a valid assumption. With these aims in mind, we first carried out a social learning experiment. For the social learning experiment, food flavour and colour were artificially manipulated in order to test whether individuals are able to learn a demonstrator individual’s food colour preference through mere observation. Results showed considerable inter-individual variability but support the assumption that bonobos can acquire and memorise information pertaining to different food preferences of group members. This experiment was followed by a playback study which is currently underway, food-associated calls from the demonstrator individuals are emitted to elicit a foraging response from listeners. Depending on the inferential abilities of listeners, caller identity and associated food preference as well as previously learned food location can inform the listener about which food is present and where. Results obtained so far are too few to draw conclusions but we will be discussed here.

February, 10th, 2016, Isaac Schamberg (animal communication)

Communication and travel coordination in wild bonobos

Simple, decentralized rules often govern patterns of group movement. Passive, local cues, however, may be insufficient to explain group behavior in all circumstances. Populations exhibiting low spatial cohesion, highly differentiated social relationships, and/or a high degree of ecological variability may be more likely to utilize explicit signals to coordinate group movement. Here, I present evidence that bonobos, a species characterized by a high degree of fission-fusion dynamics facilitate group coordination using particular call combinations and call exchanges when moving between parties. These call combinations signal relative direction of movement between parties and may be produced when more commonly observed, single call-type call bouts are insufficient to coordinate group movement.

February, 3rd, 2016, Diane Brentari

The role of prosody in the production and perception of imperatives: Data from signed and spoken languages

Imperatives stand at the crossroads between pragmatics and semantics, insofar as they are a function of several clause types and have been analyzed by some as indirect speech acts. In this study we argue that visual prosodic cues used to identify imperatives stand at the crossroads also between phonological (linguistic) and gestural (affective) categories. We pay particular attention to the visual prosodic cues that are important for identifying 4 different types of imperatives (command, explanation, advice, and permission) as they are produced and perceived in ASL and American English. The research questions posed and the results obtained are as follows:

1)   Q: Are visual prosodic features sufficient to distinguish among imperative types in perception? Our results indicate that they are sufficient to identify some imperative types in both ASL and English.

2)   Q: Do ASL and English pattern in a similar way? Our results indicate that similar visual prosodic cues for some imperative types are used in both ASL and English. Also, participants in both language groups perform similarly in perception with regard to the ease with which kinds of imperatives can be correctly identified.

3)   Q: Does audio-visual presentation in English increase performance? Our results (from this and from previous work on this topic) suggest that the addition of visual cues in English does enhance perceptual performance.

February, 3rd, 2016, Klaus Zuberbühler (animal communication)

The primate roots of human language

Human language is largely a vocal behaviour but its origins remain elusive. Vocalizations are also the main way by which primates communicate and interact socially. Sounds are produced and perceived by specialized anatomical and neural structures also present in humans. Compared to humans, however, primates are severely limited in the control they have over vocal production, which restricts their ability for rapid sound combinations and vocal learning. But language is also a cognitive capacity. Here, there is good evidence that primates understand other’s calls as given by specific individuals to specific events or social interactions. In great apes, callers take the past history with their audience into account, by suppressing, exaggerating and socially directing their calls in strategic ways. But there is no clear evidence that primates, apart from humans, perceive others as governed by mental states, nor that they are motivated to actively communicate information relevant to them.

January, 27th, 2016, Christian Rathmann

Three types of Morphology in Sign Languages

This paper uses Aronoff et al (2005) as a point of departure and examines three types of morphology in sign languages, with an eye toward understanding simultaneous morphology that changes internal properties of a sign.  One changes the sign according to fixed forms listed in the lexicon; the other looks to interaction with gestural space to determine its realization. While both are subject to language-specific constraints against marked forms, only the latter is also subject to phonological constraints against moving or twisting a manual articulator. These constraints arise because interaction with gestural space has the potential to result in forms that exceed the limits of the articulators. This latter type of nonconcatenative morphology makes sign languages unique. Furthermore, it discusses iconic properties in these morphological processes in sign languages which serve another kind of evidence for three types of morphology in sign languages

Reference:

Aronoff, Mark, Meir, Irit & Sandler, Wendy. (2005). The paradox of sign language morphology. Language, 81 (2), 301-344.

January, 20th, 2016, Tal Linzen

Are logical features encoded in distributional semantic representations?

The meaning of a content word in logical semantics is typically represented as an atom: the word "cat" means CAT. By contrast, distributional approaches represent the meaning of a word as a real-valued vector constructed from the contexts in which the word occurs. These richer representations allow us to answer new questions -- for example, we can ask whether "cat" and "dog" are more similar than "cat" and "presupposition" (the former pair is likely to share more context words than the latter).

Work on distributional semantic models typically focuses on their representation of concrete words and conceptual knowledge; it is often argued that distributional representations are unlikely to be adequate for words that are involved in logical inference, such as quantifiers, modals and negation. In practice, the boundary between the two categories is unclear; is the distinction between factive and non-factive attitude verbs conceptual or logical? More generally, do we really want two qualitatively different types of word representations?  

In this talk, I will report on an ongoing investigation of the distributional representations of words that play a central role in logical reasoning, focusing in particular on the test cases of quantifiers and attitude verbs, and testing to what extent those representations encode features such as quantificational force or factivity (spoiler: this extent will turn out to be larger than the skeptics might expect).

January, 13th, 2016, Isabelle Roy

On the development of Merge and recursion in the transition from single words to sentences

In the study of language development, children’s transition from single words to sentences remains controversial. Here we argue that Merge (Chomsky 1995) and Functional Application (FA; Heim & Kratzer 1998), driven by functional material, provide a solid basis for understanding children’s early productions. We further propose that, even though Merge can be used from the beginning of multiword productions, the ability to apply Merge in different domains develops (including the nested application of Merge, i.e., recursion). In fact, even though children use Merge correctly very early, they do not always use it in all the contexts adults do. Unlike adults, we argue, children also make use of simple juxtaposition: a conceptual rather than a grammatical operation which puts two elements together in an unordered, unstructured sequence akin to a list or a set and which cannot be associated with FA. Recourse to juxtaposition is expected to decline as facility using Merge increases to eventually reach the target grammar. Perhaps surprisingly, it is precisely the idea that not all early combinations are built by a grammatical operation, which allows us to document and analyse the development of grammatical ability in children’s productions.

December, 16th, 2015, Jeremy Kuhn

Dependent indefinites in American Sign Language

In many languages, an indefinite determiner or numeral may be inflected to indicate that the value of the indefinite DP varies with respect to another DP in the sentence or in context. In American Sign Language, inflecting the numeral ONE with an ‘arc’ movement creates such a dependent indefinite: (1) means that the books vary with respect to the boys.

(1) BOYS IX-arc-a READ ONE-arc-a BOOK.

‘The boys read one book each.’

Most semantic analyses of dependent indefinites formalize a similar insight: dependent indefinites contribute a variation condition: the value of the variable introduced by the indefinite must vary with respect to the value of another variable in the sentence or in context. The specific implementation of this insight varies in significant ways, notably on the following two fundamental architectural questions:

1. Are dependent indefinites anaphoric to their licensor, or is the relation indirect?

2. Are dependent indefinites themselves quantificational or does distribution come from a (possibly covert) distributive operator elsewhere in the sentence?

Here I argue the following: (1) dependent indefinites have an anaphoric component; (2) they are themselves quantificational. I argue that new data involving spatial agreement in ASL gives insight into these questions. An analysis will be presented within the framework of Dynamic Plural Logic (van den Berg 1996, Nouwen 2003, Brasoveanu 2008), which will be introduced during the talk.

December, 9th, 2015, Ekaterina Kravtchenko

Informationally redundant utterances trigger pragmatic inferences

Work in pragmatics shows that speakers typically avoid stating information already given in the discourse (Horn, 1984). However, it’s unclear how listeners interpret utterances which assert material that can be inferred using prior knowledge. We argue that informationally redundant utterances can trigger context-dependent implicatures, which increase utterance utility in line with listener expectations (Atlas & Levinson, 1981; Horn, 1984). In four offline experiments, we look at how comprehenders interpret utterances which refer to event sequences describing common activities (scripts, such as 'going to a grocery store').

Overall, these studies show that explicit mention of highly inferable events (e.g, 'paying the cashier,' when 'going shopping') may be systematically reconciled with an assumption that a speaker is being informative, giving rise to context-dependent implicatures regarding event ’typicality’: stating overtly that a highly typical event occurred leads to comprehenders reinterpreting it as 'atypical' in context. This effect, however, is modulated by the informational status of the utterance, possibly similar to better-known effects of prosodic focus on utterance interpretation. Overall, the results suggest that excessive informational redundancy of event utterances is perceived as anomalous, and that listeners alter their situation models in order to accommodate it.

December, 2nd, 2015, Didier Démolin (animal communication)

Grammatical structures in animal communication and the evolution of language

Comparative animal studies represent an approach to understand the evolution of phonological and syntactic features of human language. Phonology, where words are made from meaningless sounds, is considered a simpler level of combination when compared to syntax. Sound combination in animal communication raises important questions about the relevance of grammatical structures in animal communication. Recent findings in ape and monkey vocalizations showed the presence of syntactic features and therefore of grammatical structures. This is also true for several bird species. These findings open speculations about the presence, at least for apes and monkeys, of semantics in their vocalizations. They also lead us to know what is the best way to formalize these systems. The recent findings in monkey’s and bird’s systems lead us to question the uniqueness of human language among primate systems of communication in exhibiting syntax and recursion. The capacity to produce recursive, self-embedded grammars is therefore not uniquely human. These findings suggest that some essential mental processes that make human language are shared with other species.

November, 25th, 2015, Brendan Costello

Space, reference and identity: agreement in LSE (Spanish Sign Language)

Agreement in sign languages is taken to involve the use of points in space to mark the referents associated with a verb's arguments. The prevailing view in the literature has restricted agreement to a specific set of verbs, giving agreement in sign languages a somewhat haphazard flavour. Drawing on evidence from Spanish Sign Language (LSE - lengua de signos española), I propose that the basic underlying mechanism of spatial agreement occurs more broadly across the verbal domain and also in the nominal domain. In order to account for this agreement mechanism, I posit an identity ɸ-feature and examine to what extent this feature can fit into a syntactic model of the agreement process and account for the LSE data.

November, 18th, 2015, Wataru Uegaki

Ignorance and the selectional restriction of the “wonder”-type predicates

In this talk, I will discuss the semantics of “wonder" and other inquisitive predicates (e.g., “investigate", “inquire"; Karttunen 1977), and explore a solution to a long-standing puzzle about their selectional restriction, i.e., why they only embed interrogative complements and not declarative complements (except in an archaic use of “wonder", which arguably has a different meaning from “wonder"-wh). My point of departure is Ciardelli & Roelofsen’s (2015) (henceforth C&R) analysis of “wonder", which treats its lexical semantics as a conjunction of two epistemic modalities: ‘ignorance’ and ‘entertain’. According to their analysis, this conjunction is predicted to lead to a systematic contradiction iff the complement is a declarative clause. I will point out two problems with C&R’s analysis, one concerning the predicted interpretation of “wonder"-wh sentences and the other concerning the ungrammaticality of “wonder" with a ‘mixed’ complement, which involves a coordination of a declarative and an interrogative clause, as in the following.

(1) John {told me/*wonders} [which girl he likes and that he is going to ask her out].

I will argue that these two problems are solved in a unified fashion once we treat the ignorance condition of “wonder" not as mere negation of knowledge, but rather as compatibility with every alternative conveyed by the complement. I will also discuss an alternative approach to the issue of mixed complements, according to which genuine cases of mixed complements simply doesn’t exist, and (1) involves distribution of the embedding predicate over two complements (Szabolcsi 2015).

November, 4th, 2015, Steven Verheyen

A probabilistic account of vagueness

In this paper I will introduce a probabilistic account of vagueness using data from various categorization experiments in which participants were asked to decide whether or not a predicate like TALL applies to a series of stimuli (e.g., figures of various heights).  The experiments showed pronounced intra- and inter-individual differences in categorization, indicative of vague predicates. According to the proposed account these variable judgments are inherently probabilistic, with the probability determined by the distance between a stimulus’ and a participant’s position along a latent dimension. This dimension is automatically extracted from the categorization data and represents the property the participants rely on to decide on the applicability of the predicate (e.g., height in the case of TALL). A stimulus’ position on the dimension then indicates the degree to which the stimulus manifests the relevant property. A participant’s position on the dimension reflects the participant-specific degree to which stimuli have to display the property for the predicate to apply. It acts like a probabilistic threshold: the more the position of the stimulus surpasses that of the participant, the higher the probability that s/he will apply the predicate to the stimulus.

I will show how the proposed framework not only accounts for the vagueness of gradable adjectives like TALL that have a single, manifest underlying dimension (i.e., height), but also applies to vague noun categories like SCIENCES and SPORTS, with potentially multiple, latent underlying dimensions. The variable judgments for these categories too can be accounted for through the extraction of a single dimension that is best characterized as stimulus-category similarity. The framework naturally suggests the existence of two types of vagueness - vagueness in degree and vagueness in criteria - corresponding to individual differences in the positioning along the dimension of the thresholds and of the stimuli, respectively. I will show that individual differences in categorization for noun categories reflect both types of vagueness. I will conclude with a demonstration of how the framework can be used to investigate systematic sources of vagueness, by relating differences in degree (threshold positioning) and criteria (stimulus positioning) to external information about the participants, such as their age or level of education.

October, 28th, 2015, Vadim Kimmelman

Distributivity in Russian Sign Language

In this talk I discuss how distributivity is marked in Russian Sign Language. Similar to other sign languages, verbs in RSL can be marked with a distributive morpheme (in fact, with a number of different distributive morphemes with different semantics). But interestingly one of these morphemes can also mark distributivity on the arguments of the verb, and on the lexical distributive quantifier itself. I offer an agreement-based account of these facts but other possible accounts are also mentioned.

October, 21st, 2015, Alexandre Cremers & Steven Verheyen

Discussion on Gradable adjectives, vagueness, and optimal language use: A speaker-oriented model by Quing & Franke

Qing&Franke (2014) propose a Bayesian probabilistic model for relative and absolute gradable adjectives. Unlike previous model (e.g., RSA), Qing&Franke propose that the meaning of these adjectives arise as a convention among near-rational language users and does not need to be re-computed on every occasion. They also derive exact threshold for absolute adjectives.

Qing, C., & Franke, M. (2014, August). Gradable adjectives, vagueness, and optimal language use: A speaker-oriented model. In Semantics and linguistic theory (Vol. 24, pp. 23-41).

October, 14th, 2015, Amir Anvari & Mora Maldonado

Exploring redundancy

The sentence (1) is judged to be unacceptable as the whole sentence is equivalent to the first conjunct, i.e. Mary is expecting a son. A similar observation applies to the so-called Hurford disjunctions (Hurford, 1974): (2) is unacceptable, and indeed the whole sentence is equivalent to one of the disjoints, namely John was born in France. Intuitively, the deviance of these sentences is due to redundancy. However, it has been pointed in the literature (Mayr & Romoli, 2014) that redundancy is not always prohibited; for instance, (3) is exactly equivalent to (4), given that (4) is less complex than (3), we expect the latter to block the former due to redundancy, which empirically is not the case.

(1) #Mary is expecting a son and she is pregnant

(2) #John was born in Paris or France

(3) Either Mary is not pregnant or she is and John is happy

(4) Either Mary is not pregnant or John is happy

We review two theories of redundancy that attempt to explain these facts. While both theories correctly capture the deviance of (1), one, Katzir & Singh 2014, captures (2) but not (3); Mayr & Romoli 2014 on other hand can capture (3) but not (2). On top of this, we point out two data points, not previously discussed in the literature, that are problematic for both theories.

Perhaps, redundancy should be accounted for as an interaction between grammatical competence and processing mechanisms. Thus, we discuss several alternatives for studying this interaction by experimental means, starting by presenting the pilot results of an acceptability study.

October, 7th, 2015, Matic Pavlic

Word order and agreement in Slovenian Sign Language ditransitives

There is not much work carried out on sign language ditransitives although ditransitive examples regularly appear in sign language literature. From Padden (1988) on, the agree- ing verb class is usually illustrated by a ditransitive verb give. Transitive verbs may be agreeing or non-agreeing but when agreeing, they normally start in the r-locus associated with the Subject and end in the r-locus associated with the Direct Object (Od). Ditransi- tive verbs, on the other hand, are agreeing by default. They start in the r-locus associated with the Subject and end in the r-locus associated with the Indirect Object (Oi). In many SVO sign languages, ditransitives with a non-classifier predicate display SVOdOi word or- der. To my knowledge, the only study that examines word order in ditransitives with a classifier predicate is Sze (2003). She reports that they display SOdVOi word order in Honk Kong Sign Language but she does not discuss their verb-argument agreement. On a small corpus of SVO Slovenian Sign Language (SZJ) I repeat her results and go on to link ditransitive word order to the ditransitive agreement pattern. A ditransitive non-classifier predicate is signed in between the Subject and Oi; it agrees with the Subject and Oi. A ditransitive classifier predicate is signed in between Od and Oi; it agrees with Od and Oi. There is, however, one condition under which a non-regular agreement pattern may be used in SZJ non-classifier ditransitives: the subject must be first person. In this case, the ditransitive verb agrees with the Od and Oi. Crucially, the relative order of the objects may be reversed, but the agreement alignment does not change. I conclude that SZJ ditransitives are Prepositional Dative Constructions fi which implies that the Oi gets licensed by an adpositional head. I show, that in the relevant syntactic position there indeed appears a sign which could be analysed either as a functional adposition or as an agreement auxiliary.

July, 22nd, 2015, Wataru Uegaki

Veridicality and factivity of question-embedding predicates through Dayal's(1996) answerhood

This talk concerns the treatment of veridicality and factivity of interrogative-embedding predicates that also embed declarative clauses, such as "know", "tell" and "be certain". Here, we use two notions of veridicality: *veridicality with respect to declarative-embedding* and *veridicality with respect to interrogative-embedding*, defined as in (i), respectively:

(i) a. a predicate V is veridical with respect to declarative-embedding

      iff "x Vs p" entails "p"

   b. a predicate V is veridical with respect to interrogative-embedding

      iff "x Vs Q" entails, for some true answer p to Q, "x Vs p".

Based on cross-linguistic evidence, Spector & Egré (2015) put forth an empirical generalization which states that a predicate is veridical with respect to interrogative-embedding iff it is veridical with respect to declarative-embedding, contra previous claims in the literature (e.g., Karttunen 1977; Groenendijk & Stokhof 1984). This generalization is captured in Spector & Egré's analysis of question-embedding, where, roughly, "x Vs Q" can be paraphrased as "x Vs *some* answer to Q".

In this talk, I propose an alternative approach to Spector & Egré's generalization, based on the analysis where "x Vs Q" is paraphrased as "x Vs *the* answer to Q". Here, what I mean by "the answer of Q" is made precise by Dayal's (1996) answerhood operator, which presupposes the most informative true answer to Q, and returns such an answer. In this analysis, veridical (and factive) predicates, such as "know", are analyzed as evaluating the answerhood operator in the matrix evaluation world while non-veridical predicates, such as "be certain", are analyzed as evaluating the operator in worlds other than the matrix evaluation world (e.g., the subject's belief worlds). (cf. Groenendijk & Stokhof 1984)

I will discuss two aspects in which this alternative analysis has advantage over Spector & Egré's analysis. They are: (i) a precise treatment of the projection of the uniqueness presupposition in sentences involving non-veridical predicates, such as "be certain", and (ii) an analysis of factivity as the limiting case of the existential presupposition associated with Dayal's answerhood. The latter view enables solutions to two further puzzles, i.e., those concerning exhaustification above factives (cf. Klinedinst & Rothschild 2011) and an empirical generalization that factivity entails question-embeddability (modulo emotive predicates like "regret" and "resent"; cf. Egré 2008). If time permits, I would also like to compare the treatments of exhaustivity of questions embedded by "surprise"-type predicates in Spector & Egré (2015) and the current alternative approach.

July, 8th, 2015, Cory Bill

Distributive inferences in child language

A recent account of children’s behaviour with respect to scalar implicatures is based on the hypothesis that children are only able to compute inferences arising from alternatives explicitly presented to them, either as substrings of the sentence, or in the context (sometimes referred to as the Restricted Alternatives Hypothesis (RAH)). The current study tested a prediction of the RAH by investigating children’s ability to generate one type of inference that arises from alternatives that are substrings of the assertion: the so-called ‘Distributive Inference’. The results of our experiment provide tentative support the RAH, with children and adults appearing to derive DIs at similar rates, although the individual data is a bit more complicated. The implications of these results on the RAH, and other theories related to child inference generation will be discussed.

June, 24th, 2015, Thomas Geissmann (animal communication)

Gibbons: The singing apes

The territorial morning songs of the Southeast Asian gibbons or small apes (Hylobatidae) are among the most spectacular calls of mammals. They are often performed as carefully coordinated duets by mated gibbon pairs and are considered to be the best model for the evolution of human music.

Biologist Thomas Geissmann has conducted research on gibbon songs since 1980. In this talk, he will discuss functions of gibbon songs, present evidence for both inheritance of gibbon song patterns and vocal learning among singing gibbons, and compare gibbon species that use highly stereotyped song phrases with others that appear to specialize in phrases of high variability. Finally, he will discuss how far gibbon songs can be used to reconstruct gibbon phylogeny.

June, 10th, 2015, Panos Mavromatis

Chant Improvisation and the Metrical Phonology of Modern Greek

Modern Greek church chant regularly employs the adaptation of musical phrase prototypes to many different texts, and this practice often takes place in an extempore manner. Building a generative grammar of this text-tune association reveals an intimate connection between the chant's musical rhythm and the metrical phonology of the underlying text. Assuming a three-level metrical grid leads to a predictive model of melody adaptation, thus offering indirect evidence for multiple levels of stress in modern Greek beyond the lexical level.

June, 5th, 2015, Jeffrey King 

Strong contextual felicity and felicitous underspecification

Tonhauser et al [2013] and Beaver and von Fintel [2013 a,b] have argued that certain triggers for projected content put particularly stringent conditions on contexts.  They require contexts to entail their projected contents and this condition cannot be accommodated.  Tonhauser et al and Beaver and von Fintel call this condition strong contextual felicity (SCF).  Beaver and von Fintel claim that while certain contextually sensitive expressions give rise to SCF, others don’t.  They suggest that this shows that different kinds of contextually sensitive expressions have different metasemantic accounts explaining how they get semantic values in context.  I argue that all contextually sensitive expressions give rise to SCF, so that the claim that they are all governed by a single metasemantics can be reinstated.

June, 3rd, 2015, Angela Dassow (animal communication)

Exploring the Interior Structure of White-handed Gibbon Vocal Communication

Imagine that an alien landed on Earth and heard a multitude of strange sounds coming from the planet's creatures. What would it make of this? Could it distinguish species based on their sounds? Perhaps all of Earth's animals would sound alike? Could it learn these auditory signals are forms of communication? How would it decipher them? What commonalities would it find among the varied inhabitants of this world? This talk examines how this kind of scenario can be addressed. We use a combination of machine learning, formal language theory, and signal processing to analyze vocalizations from wild and captive white-handed gibbons (Hylobates lar). This approach is a radical departure from traditional studies in animal vocalization, which rely on ad hoc analyses through human observation and manual signal manipulation. By combining linguistics, computer science, and information theory, we are able to gain new insights into the communicative abilities of white-handed gibbons and demonstrate previously unrecognized complexity and structure in their vocalizations. Our approach, called Cepstral Self-Similarity Matrices, enables automatic sequencing of gibbon vocalizations into their constituent acoustic units. We analyze these sequences using basic Ergodic theory to segment them into distinct subsequences that appear consistently and repeatedly across our gibbon populations in specific referential contexts. For example, predator alarm calls share basic properties, statistical acoustic unit distributions and overall structure but each displays unique sequences associated with a particular predator. We view these as semantic units within the calls identifying the predator, as opposed to behavioral exhortations intended to trigger responses within the social group. Finally, we propose new questions which may be addressed using the above approach.

April, 29th, 2015, Hamida Demirdache

Resumption & Trace Conversion

A longstanding question in the literature on resumption is whether resumptive structures involve movement or not. The literature can be seen as dividing on lines relating to this issue. The classic view (going back to Sells 1984, for more recent proposals see Aoun, Choueiri, & Hornstein 2001, Rouveret 2002, Adger & Ramchand 2005, Guilliot 2006, Guilliot & Malkawi 2006, 2009, 2011 a.o.) is that resumptive structures do not involve movement. Over the years, however, arguments have been put forth for a movement analysis of resumption (going back to Demirdache 1991, 1997, Rouveret 1994, for more recent proposals see Aoun, Choueiri, & Hornstein 2001, Boeckx 2001, Demirdache & Percus 2008, 2011, Sechel 2012  a.o.).

This talk provides a novel argument for a movement analysis by showing how once we adopt an independently motivated view of traces —as copies with descriptive content—and, crucially, an operation for converting copy traces into definite descriptions (Trace Conversion, Fox 2002), we nicely solve a striking reconstruction puzzle with resumption in Jordanian Arabic (Guilliot & Malkawi 2006, 2009, 2011):  while resumptive epithets show all reconstruction diagnostics for movement including syntactic reconstruction (Condition C and cyclicity effects), resumptive clitics only show semantic reconstruction diagnostics (reconstruction for scope and binding).

April, 29th, 2015, Katja Guschanski (animal communication)

Phylogeny of the tribe of Cercopithecinae

Guenons (tribe Cercopithecini) are one of the most diverse groups of primates. They occupy all of sub-Saharan Africa and show great variation in ecology, behavior, and morphology. This variation led to the description of over 60 species and subspecies. Here, using next-generation DNA sequencing (NGS) in combination with targeted DNA capture, we sequenced 92 mitochondrial genomes from museum-preserved specimens as old as 117 years. We infer evolutionary relationships and estimate divergence times of almost all guenon taxa based on mitochondrial genome sequences. Using this phylogenetic framework, we infer divergence dates and reconstruct ancestral geographic ranges. We conclude that the extraordinary radiation of guenons has been a complex process driven by, among other factors, localized fluctuations of African forest cover. We find incongruences between phylogenetic trees reconstructed from mitochondrial and nuclear DNA sequences, which can be explained by either incomplete lineage sorting or hybridization. Furthermore, having produced the largest mitochondrial DNA dataset from museum specimens, we document how NGS technologies can 'unlock' museum collections, thereby helping to unravel the tree-of-life.

April, 15th, 2015, Salvador Mascarenhas

Illusory inferences: reasoning and interpretation

In this talk I explore the connection between linguistic semantics and psychology in two ways. First I argue in favor of a program for semantics as cognitive science that focuses on convergence with the psychology of reasoning.  Specifically, I show how a certain class of model-theoretic semantics (Hamblin semantics, inquisitive semantics, truth-maker semantics) can feed representations directly to a variant of mental-models theory of reasoning (Johnson-Laird, 1983), unifying insights from the two fields.  Second, I argue that the line between language-specific interpretive processes and domain-general reasoning has not been studied in sufficient detail, so that cases exist where psychology has misdiagnosed as failures of reasoning inference patterns that could in fact be the result of logically sound reasoning acting on non-obvious but legitimate interpretations of the premises. I explore both lines of research, and in particular the tension between reasoning and interpretation, taking illusory inferences from disjunction as an empirical case study.

Illusory inferences from disjunction (Walsh and Johnson-Laird, 2004) as below have been studied by psychologists focusing on the role of the domain-general reasoning module.  Over 80% of subjects in the article cited believe that conclusion (C) below follows from premises (P1) and (P2), but this is prima facie a fallacy: it could well be that Ann speaks French and Claire German, while Bill does not speak Italian.  This is the case even if we interpret "or else" in (P1) as the logician's exclusive "or" (Xor).  I give two accounts of inferences of this sort, one focusing on reasoning processes and the convergence between semantics and mental-models theory of

reasoning, the other on interpretive (pragmatic) processes.  I argue that in fact both perspectives (reasoning and interpretation) play a role in the phenomenon.  I do so by marshaling novel experimental data that extend the illusory inference paradigm by considering cases with quantifiers instead of propositional disjunction and conjunction.

(P1) Ann speaks French and Bill speaks Italian, or else Claire speaks German.

(P2) Ann speaks French.

(C) Bill speaks Italian

April, 14th, 2015, Mythili Menon

Syntactic priming across domains: Music and artificial language prime relative clauses

Theoretical   linguists   have   debated   the   relationship  between language and  music, but few psycholinguistic studies have tested whether structural  aspects  of music  prime  structural  aspects  of language. If music and language have overlapping cognitive and neural resources, then we should be able to see interference from music when participants are told to perform a music task in tandem with a language task. To test this, we conducted syntactic priming experiments where participants completed ambiguous relative clause sentence fragments in English and Spanish after listening to musical melodies. Overall, there was a main effect of musical attachment height. Low attachment musical primes resulted in a higher rate of low attachment relative clause continuations than did high attachment musical primes (p < 0.001). Language and music can activate and share abstract global representations suggesting they are similar and use shared resource networks. In sum, our results  provide striking  evidence  for  the connection between music and language at the level of cognitive processing. We will also report findings from two other experiments of similar activation and shared global representations from artificial language to natural language.

April, 8th, 2015, Carlo Cecchetto

Remarks on nominal complementation

Starting from simple facts which prima facie challenge the hypothesis that the clausal and the nominal domain have a parallel hierarchical structure, I will provide experimental evidence from two eye-tracking studies showing that a weak version of the parallelism hypothesis can be maintained only if it is assumed that nouns do not take complements.

April, 1st, 2015, Mora Maldonado

Processing scope and plural ambiguities. An experimental perspective

Our goal is to explore the psychological nature of the abstract representations and operations that are assumed to underlie semantic ambiguous sentences. We propose to use both a mouse-tracking and a priming paradigm to investigate two types of ambiguities:

1)    Scopal ambiguities: Every boy has a balloon

2)    Plural (cumulative/distributive) ambiguities: Two boys have three balloons

Priming effects can help us detect and describe the operations and representations that are actually involved in the derivation of the different readings for these ambiguities. Can we find evidence for a distributivity operator? Can we find evidence for an abstract notion of Quantifier Raising? Mouse-paths could reveal not only that one of the readings is preferred over the other one, but also whether the derivation of one reading is a necessary step for the derivation of the other.

March, 25th, 2015, Viola Schmitt

Scopelessness

My point of departure is the observation (made by Scha 1981 and many others) that DPs that may take scope w.r.t. other DPs --  here called QPs -- sometimes lack such scopal interaction. Due to the parallel of this latter construal with what we find in sentences where two or more definite plural DPs occur as co-arguments of a predicate, existing analyses usually aim to derive it qua the mechanisms standardly assumed to underlie plural predication, in particular, predicate cumulation. I argue that this approach, irrespective of its particular implementation, is flawed, my main point being that predicate cumulation does not yield the right truth-conditions for sentences with QPs. I suggest an alternative proposal,  in which the different components of standard predicate cumulation are derived in a stepwise fashion and thus become individually accessible, a point I argue to be crucial for the semantic treatment of sentences with QPs.

March, 18th, 2015, Ewan Dunbar

How much meaning can a morpheme have? Combining semantics and morphology by formally constraining the semantic learner

How much meaning can a morpheme have? Syntactic and morphological analyses generally underdetermine when distinctions in meaning between two forms are due to (i) the presence of an additional syntactic head or to (ii) different information coded on the same head. This talk presents a new semantic and morphological analysis of certain degree constructions (comparatives and superlatives: taller, tallest, more salubrious, etc) using a new formal conjecture about how this question is resolved by the learner. Surveying patterns across hundreds of languages, Bobaljik (2012) hypothesizes that superlative forms universally consist of a comparative morpheme plus an additional superlative morpheme, e.g., tallest is underlyingly [ Sup [ Cmpr [ Tall ] ] ]. Bobaljik’s hypothesis includes, in part, a speculative proposal for a universal limit on the semantic complexity of morphemes. In this talk, we offer a concrete basis for this proposal, the ‘No containment condition’ (NCC). The NCC is a constraint on language learners such that they should not consider a certain semantic representation for a hypothesized syntactic atom, if that representation can be constructed out of the semantic representations of two atoms. Illustrating the proposal, we take Bobaljik’s (2012) analysis of forms like tallest further, into [ [ Cmpr Sup ] [ Much Tall ] ]. Based in semantic analysis, our suggestion introduces Bresnan’s (1973) classical analysis of comparatives into the decomposition of superlatives.

March, 11th, 2015, Mark Steedman

A theory of content

Linguists and computational linguists have come up with some quite useful theories of the semantics of function words and the corresponding logical elements such as generalized quantifiers and negation (Woods 1968; Montague, 1973; Steedman 2012).  There has been much less progress in defining a usable semantics for content words. The effects of this deficiency are very bad: linguists find themselves in the embarrassing position of saying that the meaning of "seek" is seek'.  Computationalists find that their wide coverage parsers, which are now fast and robust enough to parse billions of words of web text, have very low precision as question answerers because, while the answers to questions like "Who wrote `The Great Gatsby?'" are out there on the web, they are not stated in the form suggested by the question, "X wrote `The Great Gatsby'", but in some other form that paraphrases or entails the answer, such as "X's `The Great Gatsby'". Semantics as we know it is not provided in a form that supports practical inference over the variety of expression we see in real text.

I'll discuss recent work with Mike Lewis which seeks to define a novel form of semantics for content words using semi-supervised machine learning methods over unlabeled text.  True paraphrases are represented by the same semantic constant.  Common-sense entailment is represented directly in the lexicon, rather than delegated to meaning postulates and theorem-proving.  The method can be applied cross-linguistically, in support of machine translation.  If I have time, I will discuss the relation of this representation of content to the prelinguistic language of mind that must underlie all natural language semantics, but which has so far proved resistant to discovery.

Lewis and Steedman, 2013a: "Combined Distributional and Logical Semantics", Transactions of the Association for Comptuational Linguistics, 1, 179-192.

Lewis and Steedman, 2014b: "Combining Formal and Distributional Models of Temporal and Intensional Semantics", Proceedings of the ACL Workshop on Semantic Parsing, Baltimore, 28-32.

March, 4th, 2015, Claire Pelofi

Ambiguity in perception and semantics in music

Pitch, which can be defined as “ that property of a sound that enables it to be ordered on a scale going from low to high ”, is crucial in many auditory processes, including the understanding of music. As a first part of this presentation, I will focus on the perception of an ambiguous pitch stimulus and present some data suggesting a main difference on the perception of ambiguity per se between musicians and non musicians.

As a second part of the talk, I will introduce a few articles addressing the question of semantics in music. I will review the main theories about emotions as a candidate for semantic content in music, focusing more particularly on the problem of culturally determined emotions in music. I will also present some findings resting upon the semantic priming paradigm, suggesting the existence of a common semantic system shared by language and music.

February, 18th, 2015, Adrien Merguerditchian (animal communication)

Les systèmes de communication des singes et des grands singes : Sur les traces des prémices du langage

Les primates non humains étant nos plus proches cousins dans l'histoire de l'évolution, étudier leurs systèmes de communication peut nous aider à déterminer les précurseurs de certaines propriétés de la parole chez nos ancêtres. Tandis que des primatologues soutiennent que la parole serait issue de l'évolution du système de communication vocale, une théorie alternative commence à gagner du terrain au sein de la communauté scientifique. Cette théorie dite "gestuelle" souligne le rôle fondamental de la communication gestuelle de nos ancêtres dans les racines évolutives du langage. Elle s’appuie sur (1) les liens entre l’organisation du langage et la communication gestuelle dans l’espèce humaine (i.e., gestes accompagnant la parole, langage des signes, gestes qui précèdent le développement de la parole chez le jeune enfant), et (2) sur les continuités mises en évidence entre le système de communication gestuelle des primates et certaines propriétés fondamentales du langage, comme l’intentionnalité, la flexibilité d’apprentissage et d’usage, les propriétés référentielles et la dominance de l'hémisphère cérébrale gauche dans le contrôle de la communication. Nos travaux et les dernières études sur les comportements communicatifs gestuels et vocaux des primates seront présentées ainsi que les données récentes en imagerie cérébrale. Ces données en éthologie, psychologie comparée et neurosciences pourraient révéler une origine non pas vocale mais "bimodale" du langage, qui serait issue de l'intégration progressive de vocalisations intentionnelles dans un système commun aux gestes communicatifs chez nos ancêtres.

February 11th, 2015, Lyn Tieu

Talk 1: Contrastive focus in children acquiring English and ASL: Cues of prominence

We show that there are common means of signaling contrastively focused elements in child English and child ASL. The first one is increase of duration of the focal element in both groups. English-speaking children marked prominence using duration (as well as other acoustic variables). ASL-signing children used this cue to prominence too, producing increased sign duration by use of sign repetition, the use of more proximal joints, and/or holds at the beginning and end of the signs accompanied by intense movement. These findings suggest that children in two linguistic modalities employ increased duration and intensity to mark salience for contrast.

Talk 2: Children's sensitivity to the ambiguity of embedded questions

Embedded questions have been argued to give rise to multiple readings, which are related in terms of strength. In this study, we investigated children’s comprehension of such questions, and found that 5-year-olds access these multiple readings. They are more tolerant, however, of weaker readings than adults. The findings have implications for theories of questions, particularly those that derive stronger readings from weaker ones through the same pragmatic enrichment that underlies the derivation of scalar implicatures. Our results are compatible with existing literature on children’s performance on scalar implicatures: in both cases children appear to be aware of the ambiguity between weaker and stronger forms, but are more tolerant of weak meanings than adults.

February, 4th, 2015, Manuel Križ

Daniel Rothschild's "A note on conditionals and restrictors"

In a recent manuscript, Rothschild investigates the common linguistic view that if-clauses denote restrictors of modal quantification in relation to two questions about conditionals in the philosophical literature: first, the way that we assign probabilities to conditionals, and second, the status of conditionals as denoting (or not) propositions. The first thesis of the paper is that some version of the restrictor view is compatible with every answer to these two questions. The second argument of the paper concerns new examples with whole conditionals inside restrictors of quantifiers, which are argued to support a propositional analysis of conditionals. We will discuss both arguments and suggest that some additional data call into question the cogency of the second in particular.

January 21st, 2015, Didier démolin (animal communication)

Source characteristics and grammatical structures in Northern muriqui spontaneous vocalizations

Is human language unique among primate systems of communication in exhibiting syntax and recursion? The recursive, hierarchical embedding of language units minimally requires a ‘context free grammar’. This is more complex than the finite state grammars thought to be sufficient to specify the structure of non-human communication signals. Animals seem unable of learning and discriminating strings of a context free grammar from those formed by simpler rules. Here we show that spontaneous combinatory calls of wild muriquis (Brachyteles hypoxanthus), a New World monkey, exhibit a grammar which is not only context-free but also context-sensitive. Thus, the capacity produce recursive, self-embedded grammars is not uniquely human. This finding suggests that some essential mental processes that make human language are shared with humans’ non-speaking cousins.

January 14th, 2015, Brigitte Senut (animal communication)

Evolution des Cercopithecoidea: les données paléontologiques

J'évoquerai l'origine potentielle des différents groupes en fonction des modifications environnementales au cours des 20 derniers millions d'années.

January 7th, 2015, Paul Egré

Vague Judgment: a Probabilistic Account

This paper investigates the idea that vague predicates like ``tall'', ``loud'' or ``expensive'' are applied based on a process of analog magnitude representation (see Fults 2009, van Rooij 2012, Solt 2012), whereby magnitudes are represented with noise. I present a probabilistic account of vague judgment, inspired from early remarks by Emile Borel (1907) on vagueness, and use it to model judgments about borderline cases. The model involves two main components:  probabilistic magnitude representation on the one hand, and a notion of subject-relative criterion. The framework is used to compare judgments of the form ``x is clearly tall'' vs. ``x is tall'', using the idea of a shift of a criterion shift. The model can be viewed as giving a naturalistic counterpart to the strict-tolerant semantics of vagueness (Cobreros et al. 2012). I then extend it to fit data concerning borderline contradictions of the form ``x is tall and not tall'' (Egré, Gardelle and Ripley 2013).

December 17th, 2014, Susan Goldin-Meadow

From homesign to sign language:  Creating language in the manual modality

Imagine a child who has never seen or heard any language at all.  Would such a child be able to invent a language on her own?  Despite what one might guess, the answer to this question is "yes".  I describe congenitally deaf children who cannot learn the spoken language that surrounds them, and have not yet been exposed to sign language, either by their hearing parents or their oral schools.  Nevertheless the children use their hands to communicate––they gesture––and those gestures, called homesigns, take on many of the forms and functions of language.  I first describe the properties of language that we find in homesign. I next consider properties of language that homesigners can and cannot develop by comparing their linguistic systems to those developed by deaf individuals in Nicaragua. Forty years ago large numbers of homesigners were brought together for the first time and Nicaraguan Sign Language (NSL) was born; NSL continues to develop as new waves of children enter the community and learn to sign from older peers. I end by taking an experimental approach to when gesture does and does not take on linguistic properties.  I examine hearing individuals asked not to speak and instead communicate using only their hands. Although these silent gesturers can create some properties of language on the spot, they do not create all of the properties that homesigners develop over time.

December 10th, 2014, Florian Schäfer

Reflexively marked anticausatives are not semantically reflexive (joint work with Margot Vivanco)

According to the standard semantics of the causative alternation along the lines of (2a, b), the truth of a causative clause (1a) entails the truth of its anticausative counterpart (1b):

(1)a. Juan  aumentó los precios   

           Juan increased the prices

 b. Los precios aumentaron(UAC)

    the prices increased

(2)a. λxλy[(y) CAUSE [BECOME [(x) higher]]]   

       b. λx[BECOME [(x) higher]]

While (1b) is an unmarked anticausative (UAC), many languages also have a set of reflexively marked anticausatives (RAC) (e.g. Spanish 3b). The SE-morpheme in (3b) is often assumed to reflect the absence of a causer argument (e.g. Grimshaw 1981, Reinhart 2000, Doron 2003, Schäfer 2008, though the technical aspects of these accounts differ lot). This general idea leaves, however, the question why the same morpheme that normally produces canonically reflexive verbs (CRV) as in (4) can serve this purpose?

(3)a. Juan rompió el  vaso

           Juan broke   the glass

       b. El vaso se rompió  (RAC)

           the glass REFL broke

(4) El  niño se lavó(CRV)

     the boy REFL washed

As an answer to this question, Koontz-Garboden (2009) (following Chierchia 1989/2004) defends the idea that the morphological identity between RACs in (3b) and CRVs in (4) reflects semantic identity (the Reflexivization Analysis of the causative alternation). In both cases, the clitic acts as a reflexivizer as in (5) that takes a transitive relation ℜ such as (6a) or (7a) as its argument and identifies the two arguments of the relation as in (6b)/(7b). The only difference between CRVs and RACs concerns the external argument θ-role: only verbs like romper select an underspecified effector (cf. 7a) lacking any agent entailments, so that the non-human theme can also be assigned this effector role: (3b/7b), under this analysis, means that 'the glass caused its own breaking'.

(5) [se] = λℜλx [ℜ(x, x)]    

(6) a. [lavar] = λxλyλe[wash(e) ∧ AGENT(e, y) ∧ PATIENT(e, x)]

     b. [se]([lavar]) =  λxλe[wash(e) ∧ AGENT(e, x) ∧ PATIENT(e, x)]

(7) a. [romper] = λxλyλsλe[∃ν[CAUSE(ν,e)∧EFFECTOR(ν,y)∧BECOME(e,s)∧THEME(s,x)∧broken(s)]]

     b. [se]([romper]) =

λxλsλe[∃ν[CAUSE(ν,e)∧EFFECTOR(ν,x)∧BECOME(e,s)∧THEME(s,x)∧broken(s)]]

We argue that RACs do not have the causative-reflexive meaning in (7b) but an inchoative one along the lines in (2b), just as UACs and other inchoative structures. Our evidence comes from the licensing of Spanish and German (meta-linguistic) negation. It follows that the SE-morpheme does not always act as a reflexivizer. We also show, however, that nothing, except world knowledge, blocks semantic reflexivization of (7a) as in (7b). The Reflexivization Analysis of the causative alternation, on the other hand, wrongly predicts meaning (2b) to be generally unavailable for RACs.

Depending on time, I will broaden our arguments to further languages and review a recent experiment by Ramchand et al. (2014) meant to support the reflexivization analysis for Norwegian RACs.

References:

Chierchia, Gennaro. 2004. A semantics for unaccusatives and its syntactic consequences. In Artemis Alexiadou et al. (eds.), The unaccusativity puzzle, 22-59. Oxford: OUP.

Grimshaw, Jane. 1981. On the lexical representation of Romance reflexive clitics. In The Mental Representation of Grammatical Relations, ed. Joan Bresnan, 87-148. Cambridge, MA: MIT Press.

Koontz-Garboden, Andrew. 2009. Anticausativization. Natural Language and Linguistic Theory 27, 77-138.

Reinhart, Tanya. 2002. The Theta System – An Overview. Theoretical Linguistics 28, 229-290.

Ramchand, Gillian, Lundquist, Bjoern & Mai Tungseth (2014): Differences in Entailment Pattern in the Causative Alternation in Germanic. Paper presented as CGSW 2014, York University.

Schäfer, Florian. 2008. The Syntax of (Anti-)Causatives. External arguments in change-of-state contexts. Amsterdam: John Benjamins.

December 3rd, 2014, Artemis Alexiadou

Language mixing and the grammar of multilingual speakers

Since the early days of modern formal approaches to grammar, most work has been based on the language of monolingual speakers. Less work has been conducted based on data from speakers who possess more than one language. Although important insights have been gained from a narrow monolingual focus, there is every reason to believe that bi- and multilingual data (henceforth, multilingual data) can inform linguistic theory. On the basis of 4 case studies involving mixing at the DP level, including bound morphemes, I will show that this is indeed the case.

November 26th, 2014, Adriano R. Lameira (animal communication)

Orangutan linguistics: new insights into the emergence of early speech components and features

Language predominantly manifests verbally, underlining every cultural and social structure in human society. Yet, language and speech evolution endure as one of the longest-standing puzzles in evolutionary sciences, partly due to the fact that no direct correspondence has ever been identified between human speech elements and great apes’ calls – our closest relatives. For instance, do ape calls correspond to what we define as vowels and consonants, syllables, words, or even brief sentences? Moreover, speech is learned anew every generation, whereas great apes seem to lack the capacity to modify, learn or imitate new calls, dividing human and nonhuman primate repertoires into two qualitatively distinct systems, seemingly deterring any possible evolutionary approach to language and speech. Recent evidence from wild and captive orangutan research indicates that fundamental similarities exist, both in basic structural components and their respective features or attributes between orangutan calls and human speech elements. Altogether, these data suggest that a new heuristic leap is eminent in the way we understand language and speech evolution, and that for the first time an innovative framework may bridge great ape ethology and human linguistics.

November 19th, 2014, Alexandre Cremers

Accounting for variable exhaustive readings, QV and homogeneity effects in embedded questions, all together.

Recent experimental results (Cremers & Chemla, 2014) confirmed the existence of several readings for sentences like (1), and we know various ways to derive them (Klinedinst & Rothschild, 2011, Spector & Egre, 2014). Quantificational variability effects arising in sentences likes (2), where the adverb seems to quantify over atomic answers to the question, have also been studied extensively since Berman (1991). However, little has been said about homogeneity effects (the fact that (3) seems to mean that John fails to know not only about one of the students, but all of them).

(1) John knows which students called.

(2) John mostly knows which students called.

(3) John doesn't know which students called.

I will present (a sketch of) a theory which aims at explaining the three effects all together. I will use a denotation of questions reminiscent of Hamblin (1973) and following Beck & Sharvit (2002), I will consider that QVE is essentially quantification over sub-questions rather than answers. Intermediate and strong exhaustive readings will be derived as global and local implicatures respectively, following Klinedinst & Rothschild (2011). However, for these two mechanisms to be compatible I will have to redefine the semantics of whether-questions in order to distinguish between weakly exhaustive and strongly exhaustive responses to a whether-question.

November 12th, 2014, Marion Dohen

Gestural (articulatory and manual) correlates of prosodic contrastive focus in oral French

It is often assumed that the production of prosodic focus solely involves fundamental frequency and intensity, thus only non visible articulators. It is also assumed that focus perception relies only on audition. The aim of this talk will be to present works showing that this is not the case. I will first present studies showing that 1. production of prosodic focus has visible articulatory correlates (production of a contrast within the utterance between what is focused and what is not); 2. these correlates play an important role in multimodal perception. I will also discuss the presence of other facial (e.g. eyebrow raising) and head movement correlates. These works are based on studies of the production of prosodic focus involving various motion capture techniques (liptracking, optotrak) as well as on several perception studies. Finally, I will also present more recent works on the joint production of prosodic focus and manual gestures (pointing and beats) especially through the analysis of hand-mouth coordination in such productions.

October 29th, 2014, Ewan Dunbar

Phonological inventories are systematic by chance: Why sound systems are shaped the way they are

Each language has a different inventory of phonemes, which we think are represented as bundles of binary features drawn from a universal repertoire ([+voice, -continuant], [-high, +back, +tense], and so on). We have known for a long time that phoneme inventories do not make fully efficient nor fully inefficient use of this binary feature system. If an inventory makes use of K features, it will not generally have a full set of 2^K different phonemes, but it will have more than K phonemes (i.e., only one phoneme for each feature contrast; Clements 2003). Put another way, inventories are neither fully systematic (each phoneme has a minimal partner phoneme along each feature dimension) nor fully asystematic.

I show that these properties do not derive from articulatory grounding of features, nor from phonology-specific Universal Grammar principles, but  match the properties of random feature matrices. I show this by summarizing all the featurally analyzed inventories in a large database of languages as binary trees showing which sets of phonemes have feature-contrasting pairs, i.e., contrastive hierarchies (Dresher 2009). I then compute a statistic over these trees to see how balanced they are, which tells us how systematic the inventory is (sum of balance, or SB statistic). Across 250 real vowel inventories, the relative prevalence of different SB scores shows an idiosyncratic pattern, with certain degrees of systematicity much more prevalent than others, but not tending towards full or zero systematicity. This confirms what we already know. However, I then show that the resulting typological tendencies match nearly perfectly what we find across 250 fake inventories, generated by uniform random sampling of binary valued matrices with their contrasts analyzed using the same contrastive hierarchy tree generating algorithm.

This strongly suggests that the reason we do not see full systematicity or full asystematicity is just that neither would be likely by chance. The patterns for symmetry or non-symmetry in the world's phoneme inventories are accidental. This is not at all the conventional wisdom in phonology; but it therefore sheds light on something which might have been thought to be a strong tendency arising from our genetic endowment (an urge toward symmetry) but which is apparently not. I finish by explaining how symmetry could still play an important role in infants' acquisition of the phonetic value of phonemic categories.

October 22nd, 2014, Heather Burnett

Reflections of Grammar in Patterns of Variation: Variable Negative Concord and the Architecture of the Syntax-Semantics Interface

This paper addresses the contributions that studies of syntactic variation can make to the construction and evaluation of formal syntactic and semantic theories. In particular, I argue that data associated with patterns of language use can provide a new testing ground for grammatical theories of phenomena at the syntax-semantics interface, one that will allow us to distinguish between existing analyses that are all descriptively adequate when applied to categorical data. As an illustration of this proposal, this work presents a new empirical investigation into asymmetric (also known as non-strict) negative concord systems of the kind found in languages like Italian, Spanish and European Portuguese. I argue that the many previous analyses of these patterns in the literature are difficult to distinguish from an empirical point of view because, despite the great diversity in their philosophical backgrounds and technical machinery, they all make roughly the same predictions when it comes to the basic paradigms associated with categorical asymmetric negative concord. I propose that one way of breaking this ‘stalemate’ is to extend the dataset that we expect our theories to account for to include variable asymmetric concord systems, such as the Montréal French system. Based on the results of a quantitative study of the distribution of bare and concord structures in the Montréal 84 spoken corpus (Thibault and Vincent, 1990), I argue that (an appropriately modified) extension of de Swart (2010)’s analysis for the Spanish categorical pattern set within a bidirectional Stochastic Optimality Theory framework provides the most satisfactory account of the forms, interpretations, and frequency distributions of sentences containing negative indefinites in Montréal 84. The success of this new analysis, which crucially relies on Jespersen (1917, 1933) and Horn (1989)’s neg first constraint, provides additional empirical support for this existence of such a principle in natural language, as well as support for the use of probabilistic grammars in the analysis of phenomena at the syntax-semantics interface. Furthermore, I argue that the case study presented in this paper opens up a new line of research that uses linguistically and socially conditioned syntactic variation data as a way to bring new empirical arguments to debates in formal syntactic and semantic theory.

REFERENCES:

Horn, L. (1989). A natural history of negation. CSLI Publications.

Jespersen, O. (1917). Negation in English and other languages. Horst, Copenhagen.

Jespersen, O. (1933). Essentials of English grammar. Allen & Unwin, London.

de Swart, H. (2010). Expression and interpretation of negation: An OT Typology. Springer, Dordrecht.

Thibault, P. and Vincent, D. (1990). Un corpus de français parlé: Montréal 84. Université Laval, Québec.

October 15th, 2014, Gillian Gallagher

Asymmetries in the representation of categorical phonotactics

An inductive learning bias in favor of constraints with the structural form of an Obligatory Contour Principle (OCP) constraint, *[αF][αF], is supported by the results of a repetition task with speakers of Cochabamba Quechua. Cochabamba Quechua exhibits two categorical phonotactic restrictions: a cooccurrence restriction on roots with pairs of ejectives *[k’ap’u], and an ordering restriction on roots with a plain stop followed by an ejective *[kap’u]. The results are consistent with a strong phonotactic restriction against cooccurring ejectives, above and beyond perception and production difficulties, while the ordering restriction seems to be represented as a weaker phonotactic restriction with speakers’ behavior primarily reflecting phonetic difficulties. As both restrictions are categorical, the results support an inductive learning bias that favors constraints like *[+cg][+cg], which penalize sequences of feature matrices with the same value for some feature, over constraints like *[-cont, -son][+cg], which penalize sequences of unrelated feature matrices.

October 8th, 2014, Dave Barner

Linguistic Alternatives and Pragmatic Inference in Language Acquisition

Pragmatic inference requires both a capacity to compute inferences and access to knowledge structures over which these inferences operate. I will argue that children's ability to compute scalar implicatures - a form of pragmatic inference - is limited by their access to relevant knowledge structures (i.e., the set of relevant scalar alternatives), and not by limitations in inference making, theory of mind, or working memory. To make this case, I will show that children's ability to make pragmatic inferences depends on the type of alternatives involved rather than the computational form of the inference (Barner & Bachrach, 2010; Barner, Brooks, & Bale, 2011; Hochstein, Bale, Fox, & Barner, 2013; Skordos & Papafragou, 2013). Also, I will show that typically developing children have the necessary epistemic reasoning abilities to compute scalar implicatures by age 5, though children with autism spectrum disorder may not. Finally, I will discuss the role of pragmatic inference and access to alternatives in the case of word learning, where lexical knowledge structures also restrict inferences regarding the meanings of new words (e.g., in domains like color, time, and number).

October 1st, 2014, Caterina Donati

The syntax of code blending: what bimodal bilinguals can teach us on language architecture

The talk analyses the code-mixing production of 6 Italian-Italian Sign Language bimodal bilinguals (Kodas)  aged 8-10 years and the grammatical judgements of adult bimodals (Codas). The data confirm  that  the  most  common  pattern  of  language  mixing  in bimodals  is  simultaneous  blending:  the  simultaneous  production  of  two language strings. I describe 5 different blending typologies classified according to the degree of autonomy of the two strings produced. The data show that very different phenomena are at play: from the full activation of only one grammar lexicalized twice selecting from the two languages, to the  simultaneous  full activation  of  two  independent grammatical representations. Important correlations and restrictions emerge, that shed light on the architecture of the bilingual competence.

July 23rd, 2014, Marion Laporte (animal communication)

The production and usage of social signals in wild chimpanzees and bonobos

Chimpanzees and bonobos are our closest cousins in evolutionary terms and in the quest to understand the origins of human language we are often tempted to attribute to their communication some features of language such as intentionality, referentiality, syntax or vocal learning. In this talk, I will first present data on the development and usage of one of the chimpanzees' most social signals,  the pant-grunt. I will also present my current work, based on the « social complexity hypothesis » for the emergence of communicative complexity which postulates that groups with complex social systems require more complex communicative systems to regulate interactions among group members. To this end, I am studying and comparing the social and vocal systems of two communities of wild chimpanzees and bonobos.

July 2nd, 2014, Berit Gehrke

Manner–in–disguise vs. intensifying WELL

It has been noted that the adverb well (and its counterparts in other languages; henceforth WELL) can function as a manner modifier, (1-a), and as a degree modifier, (1-b).

(1) a. He has written the article well. b. They are well acquainted.

While the manner reading seems to be uniformly available, we argue that what has been identified as degree WELL in the literature (e.g. Kennedy and McNally, 2005) does not correspond to a uniform phenomenon; rather, we have ‘manner-in-disguise’ WELL ((1-b)) and ‘intensifying’ WELL, which is absent in English but present in Catalan. We propose that the latter expresses the speaker’s approval of how ADJ is applied to x.

Whereas the examples to illustrate the degree reading of WELL generally involve participles, as in (1-c) (e.g. Bolinger, 1972; Kennedy and McNally, 2005), it does not seem to be possible to use WELL as a degree modifier of genuine adjectives (e.g. *the train is well blue/long/beautiful). In contrast, in Catalan, this is possible, (2).

(2) El tren és ben blau / llarg / bonic. ‘The train is pretty blue / long / beautiful.’

    the train is well blue long beautiful

This suggests that English WELL is exclusively a VP modifier (of eventualities), whereas in Catalan it can be used as an AP modifier (of adjectives or degrees), on a par with other degree modifiers such as very, rather (cp. Engl. translation of (2)).

There are further reasons to believe that the two are not the same. Kennedy and McNally (2005) point out that the ‘degree’ reading only comes about with participles associated with scales that are closed on both ends, evidenced by their compatibility with partially or fully:

(3) a. The truck is well / partially / fully loaded.

     b. ??Marge was well / partially / fully worried when she saw the flying pig.

Intensifying WELL can also combine with open scale As:

(4) ben a prop ‘WELL close’, ben amunt ‘WELL up’, ben sonat ‘WELL nuts’, ben simpatico ‘WELL kind’, ben trist ‘WELL sad’, ben viu ‘WELL alive’, ben idiota ‘WELL idiotic’

Intensifying WELL, in turn, has the following properties: It cannot be modified by other degree modifiers (5-a) (unlike well, cf. Kennedy and McNally, 2005), occur under negation, (5-b), or be questioned, (5-c) (cf. González-Rodríguez, 2006; Hernanz, 2010, for Spanish).

(5) a. En Pere és (*molt) ben simpàtic.

       the Peter is very WELL nice

b. *En Pere no és ben simpàtic. c. *En Pere és ben simpàtic?

     the Peter not is WELL nice       the Peter is WELL nice

Finally, Catalan intensifying WELL cannot be the answer to the question Com és x? ‘How is x?’, but is only felicitous when it is under discussion whether or not x is ADJ.

We assume that the adverb WELL has the same general lexical semantics as the underlying adjective GOOD (approval by some judge for something), and we follow the standard treatment of manner modifiers as predicates of event (VP modifiers); this extends to manner-in-disguise. The impression of a degree reading comes about under restricted conditions, when the event structure provides a result state associated with a closed scale that is not associated with a maximum standard (cf. McNally and Kennedy, 2013). Intensifying WELL shares the lexical semantics of good, here predicated of a property ascription, to signal that the speaker believes that x is a clear member of the set denoted by POS-ADJ; it follows that x holds ADJ-ness to a high degree. The reason why intensifying WELL is incompatible with negation and questioning has to do with a more general characteristic of degree modifiers under negation, e.g. not very ADJ is equivalent to rather un-ADJ (cf. Bolinger, 1972). With intensifying WELL, two contents are conveyed, but negation cannot target the metalinguistic one (only the high degree entailment), leading to incongruence (e.g. (5-b) would be equivalent to ‘Rather un-nice(p) & nice(p) is a good property ascription’). Finally, discourse restrictions follow from the fact that emphasis – translated here as the metalinguistic move of evaluating a property ascription – is only contextually motivated when such an ascription is challenged.

June 25th, 2014, Simon Townsend

Language evolution: syntax before phonology?

Phonology and syntax represent two layers of sound combination central to language’s expressive power. Comparative animal studies represent one approach to understand the origins of these combinatorial layers. Traditionally phonology, where meaningless sounds form words, has been considered a simpler combination than syntax and thus should be more common in animals. A linguistically-informed review of animal call sequences demonstrates that phonology in animal vocal systems is rare, whereas syntax is more widespread. In light of this and the absence of phonology in some languages, we hypothesize that syntax, present in all languages, evolved before phonology.

June 18th, 2014, Bridget Waller (animal communication)

Facial expression in human and non-human animals

Facial expression is a communication system common to all human populations. Similar facial expressions can also be seen throughout the primate order, and even in more distantly related mammals. Thus, in order to fully understand the form, social function and meaning of facial expression, we need to look beyond humans and make detailed comparisons between species. Development of anatomically based, standardised coding systems has greatly facilitated within-species and cross-species comparisons (FACS, ChimpFACS, MaqFACS, GibbonFACS, OrangFACS, DogFACS, CatFACS). These systems also minimise the risks associated with attributing emotion to facial signals a priori, which is particularly important when studying other species. Our initial studies using such tools are helping to build an evolutionary framework for a better understanding of facial expression.

June 18th, 2014, Ivy Sichel

(Resumptive) Pronouns and Competition

A Minimalist hypothesis about resumptive pronouns is that they should be no different from ordinary pronouns (McCloskey 2006). This paper substantiates the Minimalist hypothesis with respect to a particular view of ordinary pronouns: pronouns are ‘Elsewhere’ elements (Pica 1984, Burzio 1989, Safir 2004, Rooryck & Wyngaerd 2011, Reuland 2011; see also Hornstein 2001 and Grolla 2010 for resumptive pronouns). Just as the interpretation of ordinary pronouns, on this view, is determined by competition with anaphors (references above), it is argued that the interpretation of resumptive pronouns is determined by competition with gaps. Based on new facts in Hebrew I argue for the correlation between form and interpretation expressed in the Economy principle in (1).

(1) The extraction site in a Raising RC is realized by the least specified form available.

Unlike previous treatments in terms of Economy, which focused on distribution (Shlonsky 1992, Pesetsky 1998, McDaniel & Cowart 1999, Hornstein 2001, Grolla 2005, 2010), the preference is restricted to Raising relatives, and this has consequences for interpretation. The analysis is based on systematic differences between obligatory and optional pronouns.  In a variety of domains involving reconstruction and extraction, obligatory pronouns behave like gaps, whereas optional pronouns behave differently.

June 4th, 2014, Jean-Rémy Hochmann

Looking for a Speech Schema

I will review a number of experimental results suggesting that speech attracts infants’ attention from the very first days of life, and triggers dedicated learning mechanisms and neural network. These observations beg the question: how do very young infants, especially

neonates, recognize speech? What are the properties that characterize speech? Just like infants are born with a face schema, a series of properties that define what a face is, I propose that infants may be born with a speech schema, a series of properties that define what speech is. I will propose a strategy to identify this schema, comparing infants processing of speech, ape vocalizations and backward speech.

April 23rd, 2014, Uli Sauerland

Levels of Compositionality

Compositionality is often held to be a property of language rooted in human biology.  Novak et al. (2001, Nature) present a formal model of the evolution of compositionality -- i.e. a communication system with internal signal structure.  The model shows that faster learnability can be an evolutionary advantage for compositional systems. However, Novak et al.'s model doesn't consider repeated or 'mildly' compositional signals that are ubiquitous in animal communication (e.g. Arnold & Zuberbühler 2006, Curr. Biol.).  I propose that two levels of compositionality must be distinguished:  one based solely on coordination, the other allowing also subordination.  I then argue that under certain conditions Novak et al.'s learnability advantage could drive evolution from coordinating to subordinating compositionality.  As further support, I discuss evidence that subordination universally occurs in human languages.

April 16th, 2014, Jacopo Romoli

Redundancy and the notion of local context

In this talk, I discuss novel data which are problematic for Stalnaker’s (1979) non-redundancy condition, requiring not to assert something that is already presupposed. This condition has been extended to the local level, so that a sentence is deemed not assertible if it contains any part that is redundant in its local context (Fox 2008, Schlenker 2009, Singh 2007 among many others). The problem for this approach comes from disjunctions like Either Mary isn’t pregnant or (she is) and it doesn’t show. The optional presence of she is (pregnant) – a locally redundant part – is not readily predicted by the non-redundancy condition. These data are even more puzzling if compared to corresponding conditionals like If Mary is pregnant, (#she is and) it doesn’t show where the she is (pregnant) part is unacceptable as predicted by the non-redundancy condition. In response to this puzzle, we propose a solution based on Schlenker’s (2009) parsing-based theory of local contexts. In this system, exhaustifying a sentence can modify the local contexts of its parts. As a consequence of this, she is (pregnant) is actually not redundant in the disjunctive sentence above, provided the latter is exhaustified. As we discuss, this solution is not available in an approach like dynamic semantics where local contexts are computed compositionally from the syntactic structure of the sentence in question (Heim 1983, Beaver 2001; see also Chierchia 2009). Therefore, our solution to the disjunctive puzzle above, if correct, is an argument for the parsing-based approach to local contexts. More in general, redundancy provides a testing ground for these two approaches to local contexts, which are provably equivalent in the domain of presupposition projection (Schlenker 2007, 2009). We discuss also other issues that the disjunctive case above raises in connection to exhaustification, presupposition projection, and the calculation of alternatives.

April 9th, 2014, Robert May

The Conceptual Structure of Linguistic Theory

This talk will give an overview of how I see the significance of the development of generative grammar in an historical and philosophical context. Attention will be paid to the transition from structural to generative grammar, the role of semantics, and the notion of linguistic law in linguistic theory.

April 2nd, 2014, Geraci Carlo, Aristodemo Valentina and Santoro Mirko

Escaping islands: processing and morphosyntactic constraints in Italian sign language

We discuss wh-questions in complex constructions in Italian sign language (LIS) showing that:

    i) fully sentential complements are “mild” islands,

    ii) left-adjoined CPs (e.g. topicalized CPs, if-clauses, (cor)relatives, etc.) are strong islands,

    iii) right-adjoined sentential adjuncts (e.g. reason-clauses) are “mild” islands.

We claim that both processing and morphosyntactic factors conspire to determine the pattern of wh-extraction in LIS. Wh-extraction out of island domains is made possible when several processing facilitation factors are at play:

    a) d-linking;

    b) reduced linear (and possibly hierarchical) gap-filler distance,

    c) modality specific spatial and non-manual morphology.

March 19th, 2014, Marta Abrusan

Dynamic Discourse Topic

Discourse topic is a famously elusive concept. Question Under Discussion (QUD) theories (cf. Roberts 1996/2012, Ginzburg 1996, Farkas and Bruce 2010, and work in Inquisitive semantics e.g. Ciardelli et al. 2012) suggest, at least indirectly, a definition of discourse topic. We argue, however, that modeling discourse structure with QUDs is empirically inadequate for narrative texts and hence the QUD analysis of discourse topic is at best incomplete. We show that a much richer and more adequate notion of discourse topic is implicitly present in theories that model text structure on rhetorical structure, such as RST (Mann and Thompson 1988), SDRT (Asher and Lacarides 2003) or PDTB (Webber et al. 2005). Building on SDRT, we propose a notion of dynamic discourse topic that consists of the (partially ordered) set of available attachment sites at any given point in the text.  This approach not only makes more adequate predictions about topic change and topic structure in general, but also allows a uniform approach to discourse topic and anaphora resolution.

March 12th, 2014, Robert Seyfarth (animal communication)

The social origins of language

Human language is strikingly different from the communication of our closest animal relatives, the monkeys and apes, making it difficult to imagine how language could have evolved from the communication of a shared, common ancestor. The differences are clearest in call production. Continuities are more apparent, however, when one considers underlying neural mechanisms and how calls function in the daily lives of individuals. In these contexts, human and nonhuman primates share homologous brain mechanisms that have presumably evolved to serve similar social functions. In baboons – and very likely many other primates – vocalizations and social knowledge combine to form a system of communication that is discrete, combinatorial, rule-governed, and open-ended. We conclude that, long before language evolved, a discrete, combinatorial system of communication, perception, and cognition – with many of language’s supposedly unique features – was already in place.

March 5th, 2014, Amit Almor

Cross linguistic variations in the use and function of referential form

Although different languages offer different inventories of referential forms, a general principle that is believed to apply to all languages is that the most reduced form available in the language is preferred when reference is made to the most prominent entity in the discourse. In some languages, this form is an overt, phonologically realized, pronouns and in other languages this could be a null pronouns. In my talk, I will describe several studies looking at reference processing in English, Hebrew, Brazilian Portuguese and Italian that provide general support for this principles but also show some interesting exceptions. In both Spanish and Italian, speakers’ intuition as well as self-paced reading showed that null pronouns are preferred by both speakers and comprehenders. In Hebrew, a MEG study comparing the use of intra-sentential null pronouns, overt pronouns and repeated names showed a greater processing difficulty following the null pronoun than following the other forms. In Brazilian Portuguese, overt pronouns are preferred by speakers but are processed with greater difficulty than null pronouns by comprehenders. In both Hebrew and Brazilian Portuguese the dispreffered status of the null pronoun likely represents processes of language change. Together, the data from these studies suggest that while a general economical principle of balancing processing cost and discourse function can account for the general trend to prefer reduced expressions when referring to prominent discourse entities, factors related to language change, affect the processing of referential expressions. These results also suggest that the preference shown by comprehenders is not merely a reflection of usage patterns in language production. Rather, these results suggest that reference processing is a dynamic system that is characterized by stable attractors but is prone to pertubations resulting from factors external to the referential system. In my talk I will discuss the general fundamentals of this view, the novel predictions that it makes, and the challenges it faces.

February 19th, 2014, Alexander Podobryaev (MIT)

Assignment manipulation in Mishar Tatar attitude reports

Focus on novel data on indexical shifting in attitude reports in Mishar Tatar (Turkic), the discussion will be based on person indexicals that can shift and those that never can, and provide evidence that "shiftability" correlates with the possibility of being interpreted as a bound variable. An analysis of indexical shifting in terms of person-sensitive assignment manipulation (cf. Sudo 2012) capitalizes on this correlation.

An evidence for person-sensitive assignment manipulation beyond indexical shifting will be provided. Specifically, a discussion on the applicability of an analysis along these lines to some data from Collins' and Postal's "Imposters" (2012) could be done if time permits.

Feb. 11th, 2014, Florian Schwarz (University of Pennsylvania)

Experimental Investigations of Presuppositions - A Progress Report

While much of the early work in the experimental study of meaning has focused on implicatures and reference resolution, presuppositional phenomena have recently come to the forefront through work by a number of researchers. The motivation for this work has been both theoretical and processing-oriented, and the wealth of presuppositional expressions as well as their intricate interaction with the intra- and inter-sentential context still leaves many questions for future work. I take stock of some of the central existing findings and report new experimental results, primarily focusing on the time course of presupposition processing, as well as experimental perspectives on presupposition projection.

January 29th, 2014, Giorgio Magri (CNRS, U. Paris 8; UiL OTS)

The error-driven ranking model of the acquisition of phonotactics

According to the error-driven ranking model in Optimality Theory (OT), the learner starts from a restrictive initial ranking, is trained on a stream of licit forms, and instantaneously slightly reranks the constraints whenever a mistake is made on the current piece of data. This learning model has been endorsed by the OT acquisition literature (Pater and Barlow 2003, Boersma and Levelt 2001, etcetera) because it predicts a sequence of rankings that can be matched with child acquisition paths, thus modeling the observed acquisition gradualness. Yet, error-driven ranking algorithms have been dismissed by the OT computational literature (Prince & Tesar 2004, Hayes 2004, etcetera) as algorithmically too weak: the behavior of the model depends on the stream of data, so that the model feels like a leaf in the wind of data, with little guarantees on the quality of its final grammar. In particular, with little guarantees concerning the restrictiveness of the final grammar, namely its ability to correctly rule out illicit forms, despite the fact that the algorithm is only trained on licit forms. Towards a reconciliation of these two acquisition and computational perspectives, I will present some encouraging results on restrictiveness of error-driven ranking algorithms, that suggest that OT might have special formal properties that make it ideally suited in order to boost the algorithmic strength of error-driven learning.

January 22nd, 2014, Jean-Rémy Hochmann (LSCP)

Choosing different or avoiding the same?

The concepts same and different have been the topic of many studies as the paradigmatic example of abstract relations. They are also interesting for another reason: they are linked by negation. "Same" is "not different", and "different" is "not same". Do infants represent these two relations? Do they represent the logical link between them? In a series of experiments,  we found strong evidence that 7- to 14-month-old  infants can learn rules based on a representation of the relation "same", but we found no evidence of such representation for the relation "different". Infants’ behavior nevertheless suggests the existence of some representation with a content of negation. Indeed, confronted with a choice, infants learned to avoid a specific alternative. This avoidance response may constitute a precursor of the logical operator of negation. Crucially though, that negation content does not compose with the representation of "same" to produce a novel concept "different".

January 17th, 2014,  James Fuller (animal communication)

The vocal signals of adult male blue monkeys (Cercopithecus mitis): an evolutionary approach to understanding communication systems.

My research focuses, generally, on the evolution of social behavior and, more specifically, on understanding the mechanisms that maintain social relationships in group living animals. My current research examines the evolution of vocal communication systems.

Across most vertebrate taxa, vocal signals play key roles in predator avoidance, reproduction, competition, and mediating social interactions. Understanding how animals use particular signals therefore provides unique insight into species’ behavior, social dynamics, and evolution.

For the past several years, I have examined vocal behavior of adult male blue monkeys (Cercopithecus mitis) in the Kakamega Forest, Kenya. Using data from natural observation, digital audio recordings, and playback experiments, I seek to characterize the entire vocal repertoire of adult males in terms of acoustic structure, signal content (i.e. consistent relationships between signal features and attributes of signalers), and signal function.

Analyses have identified six stereotyped call types used by adult males. Each call type is acoustically distinct, yet structural similarities suggest some are more closely “related” than others. Similarly, call types exhibit distinct functions (e.g. aerial predator alarm, mate attraction) and content (e.g. social status, body size), yet evidence that some calls achieve multiple functions highlight the complexity of the repertoire. I examine these findings to infer how natural selection likely favors signal usage and how repertoires might expand over evolutionary time.

The aim of my research is to provide a comprehensive assessment of the structural and functional diversity of an entire signal repertoire, as well as insight into the socio-ecological mechanisms by which signal diversity evolves and is maintained. In doing so, I hope also to demonstrate the importance of a comprehensive approach – one that evaluates form, function, and content of entire repertoires – to understanding the use and evolution of signaling systems.

January 15th, 2014, Lucas Champollion (New York University)

Man and woman: the last obstacle for boolean coordination

The word “and” can be used both for “boolean” intersection (“John lies and cheats”) and for “non-boolean” collective formation (“John and Mary met in the park”). A major theme in the literature on coordination is the quest for a unified lexical entry. This talk argues that the boolean option is basic, focusing on conjunction in the English DP (“That liar and cheat can not be trusted”, “A man and woman met in the park”). The boolean account immediately delivers the intersective behavior of “and”. Building on seminal work by Gazdar (1980) and Winter (2001), I argue that the collective behavior falls out of its interaction with independently motivated type shifters. Essentially, "man and woman" is interpreted in a similar fashion to "a man and a woman". This talk will feature an introduction to the framework of flexible boolean semantics originally developed by Winter (2001), and time permitting, an extension of the boolean account to coordination of plurals ("ten men and women"), other languages (French: *Ce marin et soldat sont souvent ensemble), and hydras (the man and woman who met in the park).

January 8th, 2014, Lyn Tieu (LSCP)

Acquisition and learnability problems in semantics

ABSTRACT: This talk highlights the kinds of learnability problems that semantic phenomena can pose to the child learner. To demonstrate, I present experimental findings that reveal that 4-year-olds have very sophisticated knowledge about the semantics of the negative polarity item any. In particular, not only are they sensitive to licensing conditions on NPIs, they are also quite adept in their comprehension of any as a so-called domain widener. An examination of caregiver input however reveals little to no evidence for the kind of semantic mechanisms that children appear to have knowledge of (e.g., identification of subdomain alternatives, exhaustification of alternatives, domain widening). Thus I argue that in semantics, as in syntax, we are faced with the challenge of explaining children's success at language acquisition despite an apparent poverty of the stimulus. In the same vein, I will present summaries of other work in progress, examining similarly complex semantic phenomena in child language, including free choice, plurality, conjunction, and superlatives.

December 18th, 2013, Benjamin Spector (Institut Jean-Nicod)

Maximality and plural quantification - a modular approach

I will start with a puzzle regarding the semantics of modified numerals such as  `fewer than three', `between five and eight', `exactly/approximately four', etc. The puzzle is this: on the basis of their behavior in distributive environments, one is led to include in the semantics of such items a `maximality' component. For instance, the sentence `Fewer than five guests came to the party' can be paraphrased as "the *largest group* of guests who came has cardinality less than five, if such a group exists". However, in other environments, the maximality component is absent. For instance, on its cumulative reading, a sentence such as "fewer than 10 chickens managed to lay more than 100 eggs - these chickens are really quite prolific!" does not entail that the *largest* group of chickens that managed to lay (between them) more than 100 eggs has cardinality fewer than 10. Rather, it simply means that one can find a group of fewer than 10 chickens who, between them, laid more than 100 eggs (which is fully compatible with there being a larger group of chickens that have laid, between them, more than 100 eggs). After having reviewed various

conceivable ways of giving unified lexical entries to these items, I will conclude that there is no simple compositional solution to this puzzle. I will propose an alternative view where the maximality component, when present, comes from an independent mechanism, but where some general economy constraints force this mechanism to be used when a modified indefinite occurs in certain environments. I will show that this perspective allows for a new perspective on a number of problems pertaining to the meaning of words like `exactly' and `approximately'. Time permitting, I will explore how similar ideas can be applied to the semantics and pragmatics of cumulative sentences more generally.

December 18th, 2013 -  Jean Pierre Gautier (animal communication)

Diversité et Spécificité des Vocalisations des Primates

La vie en forêt, milieu optiquement dense, impose aux animaux qui y vivent des contraintes particulières dans leurs modes de communication. Sur le plan visuel, tous, présentent des livrées très colorées.  Leur mode de communication  privilégié est d’ordre sonore. On observe le développement de leur appareil phonatoire. Les primates n’échappent pas à ces deux règles,  et toutes les espèces forestières ont  pelages et des masques vivement colorés ; leurs répertoires vocaux sont tout aussi divers et remarquables, grâce au développement de leur appareil phonatoire. La diversité des vocalisations manifestées par différents groupes d’espèces, des  lémuriens aux anthropoïdes est importante.  L’exemple est pris chez les cercopithecinés, africains, cercopithèques et mangabeys.

Le répertoire vocal de ces espèces se subdivise en deux catégories : le répertoire banal, utilisés par tous les individus des groupes sociaux, et celui de mâles adultes, avec dans ce dernier le répertoire  particulier  des mâles adultes leaders sociaux de leur groupe. Il comprend toutes les vocalisations qui participent à la cohésion sociale, aux relations interindividuelles, qu’elles soient pacifiques ou agonistiques, ainsi que sexuelles. Une partie intéressante de ce répertoire comprend les différentes vocalisations d’alarme.

Le répertoire des mâles adultes, et des leaders, se différentie du répertoire banal par son extrême gravité,  sa puissance, et ses modalités d’ expression en séquences  spécifiques. La structure des ces cris, baptisés « Loud calls », montre, selon les espèces, des variations sur des aboiements,  unitaires, binaires ou en série, des  grognements ou rugissements, comme les « gobble » de mangabeys et du singe des marais. Parmi les sons les plus originaux on observe les « Booms » et le « Ooh Oup » du cercopithèque de preuss. Trois grandes catégories de cris forts existent chez les cercopithèques.  Malgré la stéréotypie des unités, on peut observer des cris de transition entre ces trois catégories, et des caractéristiques individuelles. La stéréotypie concerne aussi leurs séquences de manifestation. Celles-ci possèdent une rythmicité journalière marquée. Dans tous les cas, ces séquences ne sont jamais mono-spécifiques, mais concernent deux, trois ou quatre espèces sympatriques, avec des influences réciproques. On notera enfin que ces vocalisations possèdent une stabilité remarquable au sein de la même super-espèce, à contrario des colorations des espèces et des sous-espèces.

Ces vocalisations puissantes sont permises par la mise en fonction d’annexes extra-laryngées ou sacs vocaux.  Leur anatomie fonctionnelle sera présentée, de même que leurs niveaux de développement, selon les espèces, en liaison avec leurs performances vocales.

La structure  de vocalisations d’un individu évolue au cours de sa vie. Celle-ci peut être continue, ou discontinue. Les mâles atteignant la maturité sexuelle muent vocalement ; leur répertoire se transforme brutalement aussi lorsqu’ils atteignent la maturité sociale.

La conclusion discutera des interprétations fonctionnelles possibles, compte tenu des observations effectuées dans la nature et en captivité.

Wednesday, December 11th, 2013, Natasha Korotkova (UCLA)

Evidential shift, done by monsters

In root declarative clauses, evidentials are always speaker-oriented. Under attitude predicates, evidentials may sometimes shift, i.e. switch their perspective and become oriented towards the attitude holder. In the literature on evidentiality, it is common to derive different shifting patterns from the narrow vs. wide scope of the evidential with respect to the attitude verb. However, this view makes wrong predictions. It also overlooks that the typology of evidentials vis-a-vis their shifting properties resembles the typology of indexicals and that there are further similarities in the distribution of evidentials and shifted indexicals. Given this parallel, it seems quite natural to analyse evidential shift and indexical shift in the same fashion. Sauerland and Schenner (2007) model evidential shift after Schlenker (1999)'s account of shifted indexicals, so that the locus of cross-linguistic variation is entirely in the evidential markers. Such view has several drawbacks. In particular, it does not account for the fact that not all attitude predicates license embedded evidentials. Thus it is desirable to base an analysis of the evidential shift on a theory that has more space for cross-linguistic variation. I argue that the Monster theory (Sudo 2012) is a better alternative and provide an analysis of evidential shift within this system.

Wednesday, November 27h, 2013 - Simon Townsend (animal communication)

Syntax in non-primate animals: implications for the evolution of human language?

Human language differs from other animal communication systems in its combinatorial power. Understanding how unique this ability is to humans is crucial in unpacking how such a capacity may have evolved. To date much research has addressed call combinations in primates due to their phylogenetic proximity to humans. However, non-primate animals also represent a fruitful avenue for research given that any similarities in communicative skill between humans and non-primates can be better explained through convergent evolutionary processes rather than common descent. I provide empirical data from recent work on non-primate mammals (meerkats) and preliminary findings from birds (Chestnut crowned babblers) suggesting intriguing combinatorial abilities in the vocal domain. I will discuss the implications that these data have for our understanding of the evolutionary path that may have culminated in human language.

November 27h, 2013, Kate Arnold (animal communication)

Primate linguistics? Evidence from two species of forest primate.

November 20th, 2013, Brent Strickland (Institut Jean-Nicod)

Non-signers extract event telicity from unfamiliar sign languages: A case for non-learned mappings between visual form and abstract grammatical categories

Many of the world's languages grammatically distinguish between telic events, which encode for a logically necessary endpoint (e.g. to decide) andatelic events which could logically continue on indefinitely (e.g. to play). Previous sign language research has noted that there are regularities across many (distantly related) sign languages in the specific ways in which they encode for event telicity, with signs for telic events often involving a clear "event boundary" at the end of a gesture, while atelic events often involve repeated motion but no such clear boundary. In a series of new experiments we ask whether people who have never before been exposed to sign language are nonetheless capable of extracting information about telicity from completely unfamiliar signs. In Experiment 1, we showed native English speakers (with no previous signing experience) signs from Italian sign language. For each sign (e.g. a sign meaning "to think"), they were asked to guess the meaning, and were given a choice between the correct meaning (e.g. "to think") or a closely matched but incorrect meaning with non-matching telicity (e.g. "to decide"). Participants were at roughly 90% accuracy in guessing the correct meaning in this case. The design of Experiment 2 was similar except that both potential meanings were incorrect and semantically distant from the meaning of the original sign. However one of the two signs was matched for telicity. So if the meaning of the sign was "to play" (which is atelic), the participants would be forced to decide between "to think" (which is also atelic) and "to decide" (which is telic). In this task, participants choose the potential meaning with matching telicity at surprisingly high rates (around 70%). These results suggest that the mind may come equipped with mappings relating visual form and representations of telicity, and that such mappings may have played an important role in shaping the world's sign languages.

November 6th, 2013, Natasha Ivlieva (Institut Jean-Nicod)

Plurals and Multiplicity

The goal of the talk is to account for the behavior of bare plurals in different environments, in particular, we will focus on the following three properties:

a)    Sentences like (1) seem to imply that John saw more than one dog:

(1)                  John saw dogs on the beach.

b)    In downward-entailing contexts this “more than one” component disappears:

(2)                  John didn’t see dogs on the beach = John didn’t see one or more dogs.

c)     In the scope of another plural element, the multiplicity component does not get distributed, but there is an overall multiplicity requirement:

(3)                  My three friends attend good schools = Each of my three friends attends one or more good schools and there is more than one good school refereed to overall.

I will show that those facts can be taken care of, if we assume that plurals denote "one  or more" truth-conditionally and the multiplicity component associated with plurals is a scalar implicature (cf. Spector 2007, Zweig 2008, 2009). Moreover, I will argue that it is an obligatory scalar implicature. I will start with presenting Zweig’s account of the facts mentioned above, discuss the problems the account encounters and offer a modification designed to deal with those problems.

October 30, 2013, Manuel Kriz (U. of Vienna)

On non-maximal uses of plural definites

It is well-known that plural predications like "The professors smiled" can be used to describe situations where not absolutely all of the professors smiled (non-maximality). Another property of such sentences is that they are false only if (almost) none of the professors smiled (homogeneity). Both of these pecularities of plural predication disappear in the presence of "all". I present an analysis on which the possibility of non-maximal readings is a consequence of the interplay of the semantic property of homogeneity and pragmatic pricinples. We assume that "all", in virtue of its semantics, removes the homogeneity property from plural predication, and thereby one of the ingredients needed to derive non-maximal uses, thus explaining its precisifying effect.

October 23, 2013, Erin ZAROUKIAN

Prosody in modal concord

When responding to a question, a speaker can use rising intonation to indicate uncertainty, as demonstrated in (1a). The speaker can also include an epistemic possibility adverb like maybe in (1b). Surprisingly, though, an epistemic possibility adverb alone (with falling intonation) appears uncooperative, as shown in (1c).

(1) Amy: What is John’s favorite color?

   Ben: a. Blue?

           b. Maybe blue? ≈ (1a)

           c. #Maybe blue.

I argue that data like this demonstrates the ability of intonational operators (e.g. ?) to participate in modal concord with modal adverbs (e.g. maybe). I draw from Gunlogson (2008) on rising intonation and Anand and Brasoveanu (2010) on modal concord to provide a new analysis of rising intonation that allows for concord readings.

July 3rd, 2013, Jeremy Kuhn (NYU)

ASL Loci: Variables or Features?

American Sign Language famously disambiguates pronoun antecedents with the use of space. In ASL, both referential and quantificational DPs (e.g. Bill or every boy) can be signed at different locations ('loci') in the signing space. Pronouns can later retrieve these DPs by pointing at the same locus. Many analyses of ASL pronouns assume that these spatial loci are the overt realization of formal variables (Lillo-Martin and Klima 1990, a.o.). This assumption arises from the observation that there are arbitrarily many loci and that pronoun ambiguity can be resolved under multiple levels of embedding, mirroring the use of indices in formal systems. On the other hand, the necessity of formal variables has been contested in semantic theory; in particular, Jacobson (1999) argues for a Variable Free Semantics, grounded on the observation that variables are not logically necessary for expressive purposes.  

I present two arguments that suggest that loci should not be analyzed as variables, but rather as morphosyntactic features (as in, e.g., Neidle et al. 2000).  First, I observe a few cases in which ASL appears insensitive to accidental variable capture. I suggest that this favors a system in which features can preventthings from being coreferential, but where syntactically independent choices can’t force two pronouns to co-refer. Second, I show that loci may remain uninterpreted in certain environments (specifically, in ellipsis and under focus sensitive operators), akin to person and gender features in spoken language.

Finally, as a proof of concept, I present a variable-free fragment (using Combinatory Categorial Grammar) in which each DP carries a gradient spatial feature.  After developing an account of verbal agreement and underspecification, I show that the constraints on co-reference fall out for free under a generalized form of Jacobson's (1999) z-combinator: the syntax ensures that a pronoun and its antecedent must share the same locus.

June 26th, 2013, Giuliano Bocci (Université de Genève)

Phrase-level stress in Italian: phonological computation, information

structure and syntax.

In the first part of the talk, I present an analysis of the metrical structure of Italian. On the basis of a production experiment (Bocci & Avesani 2011), I show that postfocal material in Italian is not extraprosodic: postfocal elements, though 'given' (in sense of Schwarzschild 1999) and part of the background, are assigned phrase-level stress by virtue of default syntax-prosody mapping rules. Accordingly, I argue contra Vallduví (1991, 1992) and Szendröi (2001, 2002) that the prosodic system of Italian is not characterized by a rigid prosodic template obeying Rightmostness and that focus fronting in Italian does not result from the prosodic need to destress 'given' information. Moreover, since Italian fails to destress given information, I conclude that Givenness cannot be encoded at PF in the sense of Reinhart (2006). In the second part of the talk, I present the results of a comprehension experiment, showing the psychological reality of the proposed prosodic analysis. I discuss the consequences of the analysis for a model of prosody, and its interface with the syntax and information structure.

June 19th, 2013, Carlo Geraci (Institut Jean-Nicod)

Spatial syntax in your hands

The aim of this presentation is to argue that a specific computational component is needed in order to manage the distribution of various portions of the signing space in sign language (SL) production. This component is more easily identifiable for SLs in the sense that it is required by the visuo-gestural modality and it emerges as a condition imposed at the articulatory-perceptive interface. In standard models for language generation, like the Y model in the mainstream generative tradition [1], the faculty of language is represented as a modular system composed by a computational system, and two interpretative modules: the conceptual-intentional (CI) system and the articulatory-perceptive (AP) system [3]. Under a phase-theory approach [2], sentence-building proceeds cyclically through recursive application of the merge operation in the computational system (i.e. the core syntactic component). At the end of each phase (i.e. at Spell-out), the hierarchical structure generated by this mechanism is sent to the interpretative modules via the AP and CI interfaces. At the AP-interface several operations applies to the output of the syntactic component in order to produce the input for the phonological component. The most important ones are the linearization of the hierarchical structure, the operations connected with the morphological component and the lexical insertion. The linearization process takes the hierarchical structure generated at the end of a phase as an input and produces a linearly ordered sequence of items [4]. Concretely, this is shown by the example in (1), where the target sentence is that of a typical headfinal (SOV) language.

June 7th, 2013, Seth Yalcin (University of California, Berkeley)

"Semantics and Metasemantics in the Context of Generative Grammar"

I frame some basic foundational questions for natural language semantics as pursued in the broadly "model-theoretic" tradition informed by generative grammar.

June 5th, 2013, Luigi Rizzi (University of Siena and Chaire Blaise-Pascal)

Criteria, the labeling algorithm and the 'Halting Problem'

In the first part of the talk I will briefly present the system of Criteria for the expression of scope-discourse semantic properties: the left periphery of the clause is populated by a system of functional heads (Q, Foc, Top, etc.) which trigger A’-movement and guide interpretation at the interfaces with sound and meaning. Criterial configurations give rise to freezing effects: an element satisfying a criterion (e.g., a wh-phrase moved to the embedded C-system in an indirect question) is frozen in place and cannot be further moved (Rizzi 2006 and much related work). Issues of “further explanation” arise at this point: can freezing effects be derived from fundamental principles of linguistic computation?   

In the second part I will present Chomsky’s (2013) approach to labelling, according to which a category created by Merge is labeled by the closest head (a case of minimal search, or relativized minimality). This approach combines the criterial conception of A’-movement with elements of Moro’s (2000) dynamic antisymmetry, and offers the promise of a comprehensive account of the “Halting Problem” of A’-movement: why is stepwise movement forced to continue in certain structural environments, and forced to stop in other environments? I will propose a particular implementation of Chomsky’s algorithm and illustrate its consequences w.r.t. different kinds of freezing effects.  

April 17th, 2013, Yasutada SUDO and Benjamin SPECTOR

How Scalar Implicatures and Presuppositions Interact

March 20, 2013, Salvador Mascarenhas (New York University)

An interpretation-based account of illusory inferences from disjunction

The human capacity for reasoning is subject to failures in remarkably systematic ways.  Thus far, these phenomena have for the most part been studied by psychologists, who propose accounts rooting failures of reasoning in the general purpose reasoning mechanisms themselves.

In this talk, I discuss an alternative source for at least some of our failures of reasoning: an interpretation-based account.  On an interpretation-based view, ``failures of reasoning'' is a misnomer, at least for a representative class of attractive but fallacious inference patterns.  Rather than being the product of classically unsound general-purpose reasoning mechanisms, some of the ``mistakes'' we make in fact arise from more complex interpretive processes than meet the eye.  Specifically, I show how an important class of reasoning failures, the illusory inferences from disjunction (Walsh and Johnson-Laird, 2004), can be accounted for if we assume a classically sound reasoning module that acts upon the pragmatically strengthened meaning of premises, rather than on their literal meaning.

The account I give here is in sharp contrast with accounts of the same phenomena from psychology, which assume simple interpretive processes feeding a reasoning module specially tailored to give rise to the observed non-classical inference patterns.  The interpretation-based account also makes strong predictions, distinct from the predictions of its reasoning-based competitors from psychology.  These clear predictions allow for a novel experimental paradigm to be proposed, which aims at comparing the two classes of theories.  I show how in principle we will be able to test these predictions and decide which kind of account (reasoning-based or interpretation-based) is right for which classes of fallacies.

March 6, 2013, Brent Strickland (Yale)

Semantics may reflect "core" event structure: Language as a window into event psychology

Although semanticists and psychologists have long standing interests in the nature of event representation, interactions between these communities have unfortunately remained limited. Part of the reason is methodological. Psychologists have traditionally studied non-linguistically represented events and their influence on mental processes like attention, memory and internal time keeping. Semanticists on the other hand have (obviously) studied event representation in linguistic contexts. Despite these diverging methodologies, I argue that lessons from semantics may have important implications for psychological theories of event representation. In particular, psychology currently lacks a clear description of the computational nature of event representations (e.g. do they have internal structure or are they atomistic?), and this is exactly the type of question that work from linguistics could shed light on. In recommending a theoretical perspective capable of translating findings from one field into the other, I suggest that language may have co-opted (non-linguistic) core knowledge of events as a way of structuring semantics and the syntax/semantics interface. Thus by examining universal phenomena typically studied by linguists like thematic role assignment and telicity, psychologists can glean useful hypotheses about the structure of non-linguistically represented events. I review some related psychological studies suggesting that this perspective has already proven useful, but will also propose new lines of concrete empirical work that should be of interest to both linguists and psychologists.

February 20, 2013, Guillaume Thomas (IJN)

Tense across categories: evidence from Mbyá

In this talk, I investigate the relation between nominal tense and verbal tense in Mbyá, a Tupi Guarani language spoken in Argentina, Brazil and Paraguay, which is closely related to Paraguayan Guaraní. I argue that the functional category of tense is part both of the extended projection of the NP, and of the extended projection of the VP. I show that tense is interpreted differently in the two domains, and I offer a pragmatic explanation of this difference.

The argument focuses on the distribution and interpretation of the temporal suffixe -kue. This suffix is attested both inside NPs and in a variety of clauses (matrix, complement and relative). The nominal uses of -kue give rise to two inferences that Tonhauser (2006, 2007) identified in Paraguayan Guaraní, and which she called the Existence Property and the Change of State property. Crucially, the clausal uses of -kue do not give rise to these inferences. Tonhauser argues that the Existence and Change of State properties are built in the lexical entry of nominal -kue, and as a consequence she proposes that -kue should not be analyzed as a tense. In this talk, I explore a different analysis of -kue based on data elicited in Mbyá. I argue that the Existence Property and the Change of State Property are a mixture of local implicatures and presuppositions that arise from the interaction of the literal meaning of -kue with general constraints on the interpretation of NPs. This allows me to propose a uniform analysis of -kue across its nominal and its clausal uses, as a relative past tense. Therefore, I am able to maintain that there is a strong parallelism between the functional structure of extended VPs and that of extended NPs.

February 6, 2013, Doug Bemis (INSERM-CEA Cognitive Neuroimaging unit)

Neural signatures of simple composition: MEG evidence from basic combination during reading, speech, and non-linguistics tasks

Language derives its expressive power from the ability to combine simple elements into complex ideas. To date, however, the vast majority of neurolinguistic investigations into combinatorial language processing have focussed not on this transition from simple to complex, but rather on manipulations of complexity itself or measuring neural activity related to expectation violation. In this talk, I will present a novel paradigm that isolates neural activity related to simple compositional mechanism by combining the fine spatio-temporal resolution of MEG with the processing of minimal adjective-noun phrases (e.g. red boat). First I will demonstrate the ability of this paradigm to identify neural correlates of basic combinatorial processes that underlie the comprehension of such phrases. Then, I will present several experiments that probe the scope of these core processes both within language - comparing comprehension to production - and beyond - investigating combinatorial processing within both the pictorial and mathematical domain.

January 9, 2013, Jérémy Zehr (IJN)

Penta-valued TCS: A unified account of Vagueness and Presuppositions

Some linguistic phenomena, among which Vagueness and Presuppositions (henceforth V&P), seem to argue in favor of three-valued logics. Both V&P have been given accounts within such systems (Strong/Middle Kleene, Supervaluationism, ...) but they have never been approached in a unified way to my knowledge, though vague and  presuppositional expressions are daily used jointly in speakers' productions. The goal of the work presented in this   talk is precisely to account for speakers' judgments concerning V&P within a unified system.

First, I will focus on convergent and divergent properties of V&P. Second, I will introduce and try to adapt to presuppositions the three-valued system TCS, which has been developed to deal with vagueness and truth predicates. On the basis of the divergent properties presented before, I will thirdly extend TCS to a five-valued system that can deal with speakers' judgments in crucial contexts when V&P are involved and with propositions containing both vague and presuppositional expressions.

December 19, 2012, Vincent Homer (ENS, IJN)

What "ne...que" presupposes

(1) Paul (n’) a invité que Marie.

Paul NE has invited QUE Marie

‘Paul only invited Marie.’

Assertion: For all x other than Marie, Paul didn’t invite x.

Presupposition: Paul invited Marie.

An intuition shared by previous researchers ([1], [2], [3], [4]) is that there is hidden material in French ‘ne. . . que’ (aka NEG + EXCEPTIVE) configurations. I show that this is indeed true, and provide direct evidence that the covert ingredients in a sentence like (1) are as in (2a), i.e. (i.) an abstract negation licensing (ii.) an n-word, and (iii.) the adjective autre ‘other’ (capitals indicate hidden material):

(2) a. Paul (n’) a invité NEG PERSONNE AUTRE que Marie.

Paul NE has invited NEG anyone other QUE Marie

b. Paul (n’)a invite personne d’autre que Marie.

The adjective modifies the restrictor of the quantifier (the n-word) and takes the que-phrase as complement. The argument that is generally made in favor of postulating silent elements in a ‘ne. . . que’ configuration is that it can optionally comprise overt elements whose presence seems not to affect meaning (N.B.: in fact, they do affect the presupposition, and that is the rub). Those overt elements, the argument goes, realize covert counterparts: full realization is an option (2b); I observe that partial realization is too:

(3) a. Paul n’a NEG invité aucun AUTRE chirurgien que Marie.

b. Paul n’a pas invité de AUTRE chirurgien que Marie.

c. Paul n’a PAS invité de AUTRE chirurgien que Marie.

d. Paul n’a PAS invité d’autre chirurgien que Marie.

‘The only surgeon Paul invited was Marie.’

Actually, I show that the one element that seems to be involved in all ‘ne. . . que’ configurations is autre/AUTRE.

Now, while sharing the same truth conditions, (1) and (3) on the one hand and (2b) on the other (in the latter all the putative hidden elements that can be realized are pronounced) do not, paradoxically, have the same felicity conditions. There is a subtle yet real difference between (1)-(3) and (2b): only in the former is the prejacent presupposed:

(4) a. Paul n’a invité que Marie (=(1)), #et il n’a pas invité Marie non plus (‘and he didn’t invite Marie either’).

Presupposition of the prejacent: Paul invited Marie.

b. Paul n’a invité personne d’autre que M. (=(2b)), et il n’a pas invité M. non plus.

In sum, only the fully realized (2b) lacks the presupposition of the prejacent. Or does it? It might be, and this is the line I actually pursue, that (2b) is in fact ambiguous between two LFs, one of which carries that presupposition while the other doesn’t.

There is in fact yet another difference that singles out (2b): only in fully realized sentences can the complement of que belong to a comparison class other than the one provided by the quantifier directly modified by autre/AUTRE:

(5) a. Il n’ a rien mangé d’ autre que moi.

he NE has anything eaten of other QUE me

#‘He only ate me.’ Or: ‘He didn’t eat anything other than what I ate.’

b. #Il n’a mangé que moi.

#‘He only ate me.’

I thus submit that the two differences are related. Autre/AUTRE is a comparative adjective; I propose (i.) that it can compose either with an individual argument (phrasal comparative) or with a clausal argument and (ii.) that the insertion of silent ‘grammatical’ or ‘semi-lexical’ elements (in the sense of [6], to whose list I propose to add AUTRE, and more controversially PAS) is incompatible with the latter option. The two meanings of que moi in (5a) thus correspond to two comparative construals, the phrasal one and the clausal one respectively; only the phrasal one is available in (5b), per (ii.). I also propose (iii.) that the prejacent presupposition is triggered by and only by the phrasal comparative.

References

[1] Azoulay-Vicente, A. 1988. “La syntaxe de ne. . . que”, Lingvisticae Investigationes 12:2, 205–233. • [2] Dekydtspotter, L. P. A. 1993. “The Syntax and Semantics of the French ne que Construction”, in Proceedings of SALT 3. • [3] O’Neill, T. 2011. “The Syntax of ne. . . que Exceptives in French”, in Proceedings of the 34th Annual Penn Linguistics Colloquium. • [4] von Fintel, K. and S. Iatridou. 2007. “Anatomy of a Modal Construction”, LI, 38:3, 445–483. • [5] von Fintel, K. 1993. “Exceptive Constructions”, Natural Language Semantics 1. • [6] Kayne, R. 2003. “Silent years, silent hours”, Grammar in Focus.

December 12, 2012, Larry M. Hyman (UC Berkeley)

Why Morphology is Templatic: The Case of Bantu CARP (Causative-Applicative-Reciprocal-Passive)

In this talk I am concerned with the interplay of scope vs. morphotactics in determining suffix ordering. In Hyman (2003) I proposed a default CARP template on verbs: Causative-Applicative-Reciprocal-Passive which determines suffix ordering in Bantu. In this paper I address the following two questions (i) why has the CARP template been so “persistent” within most Bantu? (ii) what causes the dissolution of the CARP template in NW Bantu? I show that the breakdown of CARP is directly tied to innovative prosodic maximal size constraints which came to be imposed on stems (root + suffixes). As a consequence, analytic structures assume a larger role in expressing the functions of these extensions (‘cause that S’, prepositional phrases, serial verbs), even when a short verb would have offered enough room for a suffixal extension. I show that the templatic structure is not worn down by peripheral “phonetic erosion”, as generally assumed in grammaticalization studies, but rather by an interplay of prosodic constraints and restructuring from a highly agglutinative, head-marking derivational morphology to a less complex morphology and an analytic syntax. The final question addressed is: Why should languages have such templates, i.e. why should templaticity be preferred over ordering of affixes by their scope?

Reference

Hyman, Larry M. 2003. Suffix ordering in Bantu: a morphocentric approach. Yearbook of Morphology 2002, 245-281.

December 5, 2012, Guillaume Thomas (IJN)

Is the Present Tense Vacuous?

In the first part of this talk, I will reassess Sauerland's argument that the present is lexically vacuous and gets its meaning through a pragmatic competition with the past tense (henceforth, the vacuity hypothesis). This argument is based on the claim that only this hypothesis can account for the felicity conditions of sentences like (1), which can be felicitously uttered only until the last Tuesday of the month of utterance:

(1) Every Tuesday this month, I fast.

I show that once we acknowledge that such sentences are futurates, the vacuity hypothesis is no longer needed to derive the attested felicity conditions, although it does not make incorrect predictions either. Consequently, Sauerland's argument is not conclusive.

Time permitting, there will be a second part to this talk, in which I will discuss the relevance of adjunct clauses and Double Access readings of complement clauses for the evaluation of the vacuity hypothesis.

November 28, 2012, Heather Burnett (IJN)

On the Logical, Grammatical and Cognitive Foundations of Adjectival Scale Structure

In this presentation, I present a new theory of the relationship between context-sensitivity, vagueness, and adjectival scale structure. From an empirical point of view, I argue that the four principle subclasses of adjectival predicates (relative adjectives (ex. tall), total absolute adjectives (ex. dry), partial absolute adjectives (ex. wet), and non-scalar adjectives (ex. atomic)) can be distinguished along three dimensions: 1) how their criteria of application can vary depending on context; 2) how they display the characteristic properties of vague language; and 3) what the properties of their associated orders (a.k.a. scales) are. It has been known for a long

time in the literature (cf. Unger (1975), Pinkal (1995), Kennedy (2007), a.o.) that there exist connections between context-sensitivity, vagueness, and scale structure; however, a formal system that expresses these connections has yet to be developed.

By combining insights into the relationship between context-sensitivity and scalarity from the delineation semantics framework (Klein (1980), a.o.) with insights into the relationship between tolerance relations and the Sorites paradox from Cobreros, Égré, Ripley & van Rooij (2012)’s Tolerant, Classical, Strict (TCS) framework, I propose such a logical system. Using this framework, I show that the association of particular classes of adjectives with their particular kinds of scales can be derived from their context-sensitivity and vagueness properties. In other words, I argue that from independently necessary theories of context-sensitivity and vagueness, we arrive at a full theory of gradability and scale structure in the adjectival domain.

October 24, 2012, Guillaume Thomas (IJN)

Embedded Imperatives in Mbyá

It has often been claimed that imperatives cannot be embedded. This claim has been challenged recently with datas from a variety of languages (Rus 2005 on Slovenian, Chen-Main 2005 on Mandarin, Crnic and Trinh 2009 on English). In this talk, I investigate imperatives embedded under verbs of report and request and under a reportative evidential in Mbyá Guaraní, a Tupí-Guaraní language spoken in Argentina, Brazil and Paraguay. Using Krifka's (to appear) recent work on embedded speech acts, I propose that imperatives are a clause type that is associated with directive speech acts and that operators that embedded imperatives are either illocutionary modifiers or verbs that describe speech acts. I consider cross-linguistic predictions of the analysis.

October 3, 2012, Yasutada Sudo (IJN)

Non-entailed Presuppositions and Projection in Quantified Sentences

Presuppositions of quantified sentences are known as a recalcitrant problem for theories of presupposition projection. In this talk I will first show that certain kinds of presuppositions are not entailed by assertive meanings, and that they pose a serious challenge for existing theories of presupposition projection. Then I propose a new theory that accounts for the projection properties of various quantificational determiners in a principled way.

September 26, 2012, Roumyana Pancheva (USC, Los Angeles)

NUMBER, many, more and most

The relation between proportional 'most' (as in “Mary read most articles.”) and superlative 'most' (as in “Mary read the most articles.”) has been of great interest (see Szabolcsi 2010 for discussion). Recent accounts provide compositional analyses of proportional 'most' as a syntactically complex superlative expression (Hackl 2009, Gajewski 2011, Szabolcsi 2012). There are languages, however, that have a superlative 'most' but not a proportional 'most'. I suggest that this fact does not threaten the compositional treatment of proportional 'most'; rather, it reveals the even more complex structures behind the two cardinality 'most's, involving null NUMBER (as in Kayne 2005), the ‘Q’-adjective many, and their comparative and superlative forms.

September 19, 2012, Roumyana Pancheva (USC, Los Angeles)

Superlatives, Focus and Scope

June 6, 2012, Andrea Moro (IUSS, Pavia)

Dynamic Antisymmetry; Movement as Geometrical Effect

Movement is a specific property of human languages and one that has at least implicitly been recognized in all linguistic theories. Certain (groups of) words appear to be interpreted in a different position with respect to the one in which they occur in the sequence building the expression. For example,  "which books"  in "which books did John know that Mary read?" is interpreted as if it were in the object position of the verb "read" in the embedded sentence. This property is generally considered as a consequence of morphological requirements of grammars. In this paper an alternative approach will be proposed according to which displacement is the consequence of process of linearization/computation underlying the recursive architecture of human syntax.

April 18, 2012, Kyle Duarte (IJN)

Phonetics and Phonology for Signing Avatars

The creation of signing avatars (i.e., virtual characters that use signed languages) has become a popular research topic in recent years, but how they sign still feels robotic or disinterested to native signers. The work I have performed for my PhD dissertation aims at improving signing avatars to better target native users by 1) developing a focused database of signs that can be used for animation, 2) performing linguistic analyses on the data to inform animation techniques, 3) animating sequences for the purposes of evaluation, and 4) evaluating the effectiveness of our animation system at increasing the rapport between avatars and their interlocutors. This work is done in an iterative fashion with increasingly finer detail to ensure that issues are identified and addressed as they become apparent. Some issues remain to be addressed, however - while this work is exciting for linguists and computer scientists alike, it needs to be done with involvement from the Deaf community if it is to be adopted by and provide direct benefit to its users.

March 28, 2012, Michael Franke (University of Amsterdam)

Scalar Items in Embedded Position: An(other) Experimental Approach

If a scalar item occurs in the scope of another logical operator, usually several types of readings are conceivable. For instance, an utterance of sentence (1) could allow for either a literal (2a), a globally strengthened (2b) or a locally strengthened reading (2c).

(1) All of the students read some of the papers.

(2) a. All of the students read some or all of the papers.

    b. All of the students read some of the papers, and it's not the

      case that all of the students read all of the papers.

    c. All of the students read some but not all of the papers.

In particular the existence of locally strengthened readings (2c) is a matter of debate in the theoretical literature (e.g., Chierchia, Fox and Spector 2008, Geurts 2010). Unfortunately, recent experimental work, if taken in conjunction, has also only provided inconclusive evidence here (Geurts & Pouscoulous 2009, Chemla & Spector 2011).

We suggest that previous studies might be insufficiently informative because of the focus on (variants of) a picture-verification paradigm. The problem is that in order to test the availability of different candidate meanings different pictures have to be presented, so that effects of pictorial complexity or stereotypicality could never be ruled out entirely. We therefore studied a different mode of visual presentation: subjects were presented with pictures that were partly covered and could incrementally be uncovered at the subjects' request; for each ((partially) covered) picture, subjects had to decide whether they could already give a truth-value judgement or needed more information. Moreover, since it is often argued that intonational stress on an embedded scalar item can favor a locally strengthened reading we presented sentence auditorily and manipulated stress accordingly.

Preliminary inspection of our data shows hardly any evidence for globally strengthened readings, but some evidence for locally strengthened readings. Contrary to claims in the theoretical literature, we found no evidence that stress favors locally strengthened readings.

Februrary 22, 2012, Isabelle Charnavel (École normale supérieure and UCLA)

Binding and Focus: What propre Reveals

Crosslinguistically, it is largely attested that intensified pronouns such ashimself typically become anaphoric. Although this is discussed especially in the typological literature, it remains unclear exactly how. From a compositional standpoint, this difference must be attributed to the presence of self. In this talk, I show how the behavior of the French adjective propre in expressions such as sa propre mère (very similar to that of own in her own mother) sheds light on this question by demonstrating the existence of a synchronic interaction between binding and focus and thus suggesting that Binding Theory is deeply connected with the syntax and semantics of Focus. In addition, it further demonstrates the direct connection between anaphoric and logophoric behaviors of reflexive-like expressions crosslinguistically.  Indeed, the binding behavior of expressions such as [son propre père] depends on which of two main focus-based interpretations is associated with propre: when propreinduces alternatives to the possessor, son behaves either like an anaphor or like a logophor (depending as we will see on the animacy of the antecedent); but when propre induces alternatives to the possessum, sondoes not obey any binding requirements and instead induces scalarity effects in the sense that the possessum corresponds to a least expected alternative.  The long term goal is of course to elucidate this synchronic link between binding and focus. In this talk, I present intermediate results in this direction. First, I show that the behavior of possessor son (propre)provides insights into the relevant notions to be used for Binding Theory: it rehabilitates the classical binding theory against coargument views of binding in showing the relevance of syntactic local domains rather than coargumenthood; moreover, it argues in favor of a theory of exemption from the narrow binding theory (Condition A) related to logophoricity. Next, I demonstrate that the only way to derive the possessum interpretation of son propre - i.e. to predict the right domain for the scalarity effect induced by son propre - is to postulate the existence of a covert, focus sensitive operator E close to overt even; this argues in favor of theories of covert focus sensitive operators and more generally for the necessity to postulate abstract syntactic representations.  To conclude, I explore how these results could yield a deeper understanding of the link between focus and binding revealed by the behavior of propre.

February 15, 2012, Christina Pawlowitsch (Ecole d’Economie de Paris)

Neutral stability, drift, and the diversification of languages

The diversification of languages is one of the most interesting facts about language that seek explanation from an evolutionary point of view. Conceptually the question is related to explaining mechanisms of speciation. An argument that prominently figures in evolutionary accounts of language diversification is that it serves the formation of group markers which help to enhance in-group  cooperation. In this paper we use the theory of evolutionary games to show that language diversification on the level of the meaning of lexical items can come about in a perfectly cooperative world solely as a result of the effects of frequency-dependent selection. Importantly, our argument does not rely on some stipulated function of language diversification in some coevolutionary process, but comes about as an endogenous feature of the model. The model that we propose is an evolutionary language game in the style of Nowak et al. [1999, The evolutionary language game. J. Theor. Biol. 200,  147--162], which has been used to explain the rise of a signaling system or protolanguage from a prelinguistic environment. Our analysis focuses on the existence of neutrally stable polymorphisms in this model, where, on the level of the population, a signal can be used for more than one concept or a concept can be inferred by more than one signal. Specifically, such states cannot be invaded by a mutation for bidirectionality, that is, a mutation that tries to resolve the existing ambiguity by linking each concept to exactly one signal in a bijective way. However, such states are not resistant against drift between the selectively neutral variants that are present in such a state. Neutral drift can be a pathway for a mutation for bidirectionality that was blocked before but that finally will take over the  population. Different directions of neutral drift open the door for a mutation for bidirectionality to appear on different resident types. This mechanism---which can be seen as a form of  shifting balance---can explain why a word can acquire a different meaning in two languages that go back to the same common ancestral language, thereby contributing to the splitting of these two languages. Examples from currently spoken languages, for instance, English clean and its German cognate klein with the meaning of ''small,'' are provided