HU

Itamar Kastner

Papers and Talks

E-mail me if you'd like the TeX source for any of these. See my CV for the full list, including invited talks and papers.

Manuscripts

Under reviewItamar Kastner. Templatic morphology as an emergent property: Roots and functional heads in Hebrew.
[LingBuzz] [abstract +/-]

Modern Hebrew exhibits a non-concatenative morphology of consonantal "roots" and melodic "templates" that is typical of Semitic languages. Even though this kind of non-concatenative morphology is well known, it is only partly understood. In particular, theories differ in what counts as a morpheme: the root, the template, both, or neither. Accordingly, theories differ as to what representations learners must posit and what processes generate the eventual surface forms. In this paper I present a theory of morphology and allomorphy that combines lexical roots with syntactic functional heads, improving on previous analysis of root-and-pattern morphology. Verbal templates are here argued to emerge from the combination of syntactic elements, constrained by the general phonology of the language, rather than from some inherent difference between Semitic morphology and that of other languages. This way of generating morphological structure fleshes out a theory of morphophonological alternations that are non-adjacent on the surface but are local underlyingly; with these tools it is possible to identify where lexical exceptionality shows its effects and how it is reigned in by the grammar. The Semitic root is thus analogous to lexical roots in other languages, storing idiosyncratic phonological and semantic information but obeying the syntactic structure in which it is embedded.
Under reviewAn MEG experiment on morphological decomposition and lexical access in Hebrew, and an MEG experiment on production in ASL. E-mail me for details if you're interested.
2016Itamar Kastner. Form and Meaning in the Hebrew Verb. Doctoral dissertation, New York University.
[LingBuzz] [pdf] [abstract +/-]

This dissertation is about the basic building blocks that make up words, and how these building blocks interact with the rest of the grammar. The grammar is generally viewed as an inventory of contentful units and the rules governing their combination. One question for linguistic theory is what these units might be like. Are they different for different languages? A second question is how these pieces are put together, and again we ask whether these combinatory processes are the same in different languages. A third question is to ask what we can build. This study concentrates on building verbs, specifically how the grammar builds their structure in a way that then constrains semantic interpretation and phonological pronunciation.
The empirical domain is the verbal system of Modern Hebrew, where this work attempts to unify our treatment of concatenative and non-concatenative morphology. The hypothesis put forward is that hierarchical syntactic structure, once generated, must be interpreted according to specific locality constraints when transferred to the interfaces with semantics and phonology. At each interface additional calculations take place. These calculations are interface-particular: semantics and phonology are not identical objects of study. Yet the two have in common a locality constraint on calculations that derives directly from the syntactic structure. In addition, individual lexical items ("roots") place their own requirements on the meaning and/or the pronunciation. The theory developed here limits this influence of roots to the two interfaces, making the claim that individual roots have no syntactic features. Nevertheless, roots are active at the interfaces in ways that are predictable once the right generalizations are sought out. The phonological form of roots is relevant at the phonology and their lexical semantics is relevant at the semantics: roots have no syntactic features, only interface requirements.
Hebrew, being a contemporary Central Semitic language, shows the kind of non-concatenative, "root-and-pattern" morphology that is organized around consonantal "roots" and prosodic "templates", the latter consisting of a prosodic shape, certain vowels and an affix. The account put forward argues that Hebrew roots are abstract lexical elements which combine with discrete syntactic functional heads. The combination, once fed through the phonology of the language, results in morphophonological templates which are not primitives of the system in and of themselves. The architecture defended supports the view of constrained interpretation at the interfaces which lies at the core of this proposal, using non-hierarchical surface forms in order to mount an argument for hierarchical structure.
Chapter one of the dissertation introduces the issues at hand and the basics of the Hebrew verbal system. It also reviews a number of earlier approaches which help set the stage for the analysis that follows.
Chapter two develops the syntactic-semantic part of the proposal, defining the syntactic elements needed to derive verbal morphology both for Hebrew and crosslinguistically. It is shown that the different combinations of these elements produce the verbal system of Hebrew in a way that is constrained, in the semantics, by the lexical idiosyncrasies of individual roots.
Chapter three takes the proposed structures and manipulates them in the phonological component of the grammar. The view of linearization pursued here is shown to make correct predictions. The effect of different classes of roots is highlighted, and the point is made that verbal templates are not holistic morphemes but the spell-out of distinct functional heads.
Chapter four takes a quantitative approach, surveying previous psycholinguistic and neurolinguistic work on Semitic languages and presenting novel findings from a recent magnetoencephalography experiment. These findings support the claims made in the previous chapters regarding the organization of the system.
Chapter five considers how the child might acquire this system. Recent developmental findings are surveyed and a novel computational model is discussed. This chapter outlines a model of Semitic acquisition in which the consonantal character of roots is used as a learning cue, leading to acquisition of basic verbal templates and eventually the system as a whole.
Chapter six concludes, recapitulating the main contributions of this work: derivations in a generative grammar combine rigid grammatical principles with unstructured lexical material. This dissertation defends an explicit view of how such combination takes place.
2014Itamar Kastner and Vera Zu. The syntax of implicit arguments. Ms., NYU.
[info +/-]
A term paper from 2014 that Vera and I get asked for sometimes. E-mail me for a copy.

Articles

2017Itamar Kastner. Reflexive verbs in Hebrew: Deep unaccusativity meets lexical semantics. Glossa.
[paper] [abstract +/-]

Reflexive verbs in Modern Hebrew show specific morphological marking: only one of the seven verbal templates in the language can be used for reflexives. Yet this morphological marking also appears on anticausative and reciprocal verbs, which have different syntactic and semantic properties. I provide an analysis of reflexivity in Hebrew which does not make reference to dedicated reflexive morphosyntax. By combining independent functional heads, the proposal explains what in the syntax underlies this morphology and how different kinds of verbs end up with identical morphophonological properties. To this end, I consider the lexical semantics of individual lexical roots as well as the syntactic configurations in which roots and arguments are embedded. The resulting theory is one in which lexical roots trigger specific interpretations of the syntax at the interfaces.
2017Itamar Kastner and Frans Adriaans. Linguistic constraints on statistical word segmentation: the role of consonants in Arabic and English. Cognitive Science.
[paper] [preprint] [abstract +/-]

Statistical learning is often taken to lie at the heart of many cognitive tasks, including the acquisition of language. One particular task in which probabilistic models have achieved considerable success is the segmentation of speech into words. However, these models have mostly been tested against English data, and as a result little is known about how a statistical learning mechanism copes with input regularities that arise from the structural properties of different languages. This study focuses on statistical word segmentation in Arabic, a Semitic language in which words are built around consonantal roots. We hypothesize that segmentation in such languages is facilitated by tracking consonant distributions independently from intervening vowels. Previous studies have shown that human learners can track consonant probabilities across intervening vowels in artificial languages, but it is unknown to what extent this ability would be beneficial in the segmentation of natural language. We assessed the performance of a Bayesian segmentation model on English and Arabic, comparing consonant-only representations with full representations. In addition, we examined to what extent structurally different proto-lexicons reflect adult language. The results suggest that for a child learning a Semitic language, separating consonants from vowels is beneficial for segmentation. These findings indicate that probabilistic models require appropriate linguistic representations in order to effectively meet the challenges of language acquisition.
2017Itamar Kastner and Vera Zu. Blocking and paradigm gaps. Morphology.
[paper] [preprint] [abstract +/-]

Gaps in morphological paradigms are often explained in terms of blocking: generating one form is blocked by the existence of a paraphrase. Another way of thinking about paradigm gaps dissociates their existence from competition between forms. Unlike in competition-based approaches, systematic gaps can be seen as true gaps; the system might not generate a certain form, but this form is not considered in comparison to others. Adopting this latter approach, we argue that inflectional paradigms are neither morphosyntactic primitives nor the result of competition. This claim is supported by data from two unrelated languages. For Hebrew, we demonstrate that a passive gap is not the result of competition with analytic paraphrases. For Latin, we show that a cyclic, syntax-based approach is superior to a theory that generates nonactive verbs in the lexicon and has them compete against each other. Systematic paradigm gaps are thus argued to result from syntactic structure building, without competition regulating independent morphological constructions.
2017Irit Meir, Mark Aronoff, Calle Börstell, So-One Hwang, Deniz Ilkbasaran, Itamar Kastner, Ryan Lepic, Adi Lifshitz Ben Basat, Carol Padden and Wendy Sandler. The effect of being human and the basis of grammatical word order: insights from novel communication systems and young sign languages. Cognition 158:189-207.
[paper] [abstract +/-]
This study identifies a central factor that gives rise to the different word orders found in the world’s languages. In the last decade, a new window on this long-standing question has been provided by data from young sign languages and invented gesture systems. Previous work has assumed that word order in both invented gesture systems and young sign languages is driven by the need to encode the semantic/syntactic roles of the verb’s arguments. Based on the responses of six groups of participants, three groups of hearing participants who invented a gestural system on the spot, and three groups of signers of relatively young sign languages, we identify a major factor in determining word order in the production of utterances in novel and young communication systems, not suggested by previous accounts, namely the salience of the arguments in terms of their human/animacy properties: human arguments are introduced before inanimate arguments (‘human first’). This conclusion is based on the difference in word order patterns found between responses to depicted simple events that vary as to whether both subject and object are human or whether the subject is human and the object inanimate. We argue that these differential patterns can be accounted for uniformly by the ‘human first’ principle. Our analysis accounts for the prevalence of SOV order in clauses with an inanimate object in all groups (replicating results of previous separate studies of deaf signers and hearing gesturers) and the prevalence of both SOV and OSV in clauses with a human object elicited from the three groups of participants who have the least interference from another linguistic system (nonliterate deaf signers who have had little or no exposure to another language). It also provides an explanation for the basic status of SOV order suggested by other studies, as well as the scarcity of the OSV order in languages of the world, despite its appearance in novel communication systems. The broadest implication of this study is that the basic cognitive distinction between humans and inanimate entities is a crucial factor in setting the wheels of word ordering in motion.
2015Itamar Kastner. Factivity mirrors interpretation: The selectional requirements of presuppositional verbs. Lingua 164:156-188.
[paper] [preprint] [abstract +/-]

Different verbs can take different kinds of arguments. Factive verbs such as remember and forget take clausal complements which are presupposed to be true. Verbs such as say and think do not presuppose the truth of their complements. I suggest that the complements of presuppositional verbs like remember, forget, admit and deny are actually definite DPs, picking out a discourse referent in the Common Ground. It has been established that certain effects arise in clauses embedded by presuppositional verbs: only complements can be extracted from them and argument fronting inside of them is not allowed. We also note that they can have DP pro-forms (which). This stands in opposition to non-presuppositional clauses, where extraction and fronting are possible, and only CP pro-forms (as, so) are possible. The DP view allows for a uniform solution to these puzzles, based on spoken and signed languages; the crucial distinction is whether the matrix verb selects for a definite entity (presuppositional) or for a proposition (nonpresuppositional). Finally, we introduce crosslinguistic data showing that the entity/proposition split parallels a split in interpretation between DP and CP complements: a presuppositional DP complement is interpreted like a presupposed entity, while a nonpresuppositional CP complement has the semantics of a novel proposition. This way of looking at clausal complements also allows us to account for the behavior of sentential subjects, which have been argued to be both nominal and factive. I flesh out these generalizations and show that they emerge as a natural result in our framework. With sentential subjects as with clausal complements, factivity and presupposition correlate with the syntactic category of the argument. I take this to imply a form-meaning isomorphism in the syntax and semantics of what a verb licenses (DP vs CP).
2015Masha Westerlund, Itamar Kastner, Meera Al Kaabi and Liina Pylkkänen. The LATL as locus of composition: MEG evidence from English and Arabic. Brain and Language 141:124-134.
[paper] [abstract +/-] [highlights +/-]

Neurolinguistic investigations into the processing of structured sentences as well as simple adjective-noun phrases point to the left anterior temporal lobe (LATL) as a leading candidate for basic linguistic composition. Here, we characterized the combinatory profile of the LATL over a variety of syntactic and semantic environments, and across two languages, English and Arabic. The contribution of the LATL was investigated across two types of composition: the optional modification of a predicate (modification) and the satisfaction of a predicate’s argument position (argument saturation). Target words were presented during MEG recordings, either in combinatory contexts (e.g. “eats meat”) or in non-combinatory contexts (preceded by an unpronounceable consonant string, e.g. “xqkr meat”). Across both languages, the LATL showed increased responses to words in combinatory contexts, an effect that was robust to composition type and word order. Together with related findings, these results solidify the role of the LATL in basic semantic composition.

- We investigated the generality of linguistic composition responses in the LATL.
- We measured responses to two types of composition: argument saturation and modification.
- We measured responses in two languages, English and Arabic.
- The LATL is robustly engaged in composition across both rule types.
- The LATL shows similar combinatory responses across both languages.
2014Itamar Kastner, Irit Meir, Wendy Sandler and Svetlana Dachkovsky. The Emergence of Embedded Structure: Insights from Kafr Qasem Sign Language. Frontiers in Psychology 5:525.
[paper] [abstract +/-]

This paper introduces data from Kafr Qasem Sign Language, an as-yet undescribed sign language, and identifies the earliest indications of embedding in this young language. Using semantic and prosodic criteria, we identify predicates that form a constituent with a noun, functionally modifying it. We analyze these structures as instances of embedded predicates, exhibiting what can be regarded as very early stages in the development of subordinate constructions, and argue that these structures may bear directly on questions about the development of embedding and subordination in language in general. Deutscher (2009) argues persuasively that nominalization of a verb is the first step -- and the crucial step -- towards syntactic embedding. It has also been suggested that prosodic marking may precede syntactic marking of embedding (Mithun 2009). However, the relevant data from the stage at which embedding first emerges have not previously been available. Kafr Qasem Sign Language might be the missing piece of the puzzle: a language in which a noun can be modified by an additional predicate, forming a proposition within a proposition, sustained entirely by prosodic means.
2009 Itamar Kastner and Christof Monz. Automatic Single-Document Key Fact Extraction from Newswire Articles. In Proceedings of the Conference of the European Chapter of the Association for Computational Linguistics (EACL 2009), pages 415-423.
[pdf] [bib] [data] [slides] [abstract +/-]

This paper addresses the problem of extracting the most important facts from a news article. Our approach uses syntactic, semantic, and general statistical features to identify the most important sentences in a document. The importance of the individual features is estimated using generalized iterative scaling methods trained on an annotated newswire corpus. The performance of our approach is evaluated against 300 unseen news articles and shows that use of these features results in statistically significant improvements over a provenly robust baseline, as measured using metrics such as precision, recall and ROUGE.
Book review
(2016)
Review of The Syntax of Roots and the Roots of Syntax (Alexiadou, Borer and Schäfer 2014, Oxford University Press). Language 92(1):210-212.
[link]

Conferences

2017 Itamar Kastner and Tal Linzen. A morphosyntactic inductive bias in artificial language learning. In the 48th annual meeting of the North East Linguistic Society (NELS 48).
[abstract]
Itamar Kastner. Revisiting the roots of Semitic morphology. Roots V, Queen Mary University of London.
[abstract]
Yohei Oseki and Itamar Kastner. The trivalency of Voice. The Cambridge Workshop on Voice.
[handout]
Itamar Kastner. Coercing propositional anaphora. Workshop on Polysemy and coercion of clause-embedding predicates, DGfS 39, Saarbrücken.
[slides]
2016 Itamar Kastner. Passives in non-concatenative morphology. Genus Verbi, Wittenberg.
Itamar Kastner. Impossible morphemes: The case of Semitic templates. Generative Grammatik des Südens 44, Leipzig.
[info +/-]
A short version of my "emergent property" paper.
Itamar Kastner. The zero-derived causative alternation in Hebrew is rare, but systematic. Workshop on The Word and the Morpheme, Humboldt University.
[abstract]
Itamar Kastner. A non-uniform account of intransitive verbal forms in Hebrew. In the 90th annual meeting of the Linguistic Society of America (LSA).
[info +/-] [handout]

Superseded by Chapter 2 of my dissertation.
2015 Itamar Kastner and Frans Adriaans. Consonant representations aid in learning segmentation and phonology for Arabic but not English. Poster presentation at the 40th Boston University Conference on Language Development (BUCLD 40).
[info +/-] [poster] [abstract]

Superseded by Kastner and Adriaans (2017).
Itamar Kastner and Frans Adriaans. Language-specific representations: learning segmentation and coocurrence restrictions in Arabic. In the 9th North East Computational Phonology Circle (NECPhon 2015).
[info +/-]

Superseded by Kastner and Adriaans (2017).
Itamar Kastner. Syntactic versus phonological generalizations in Hebrew verbal morphology. In the 46th annual meeting of the North East Linguistic Society (NELS 46).
[info +/-] [poster] [abstract]

In this NELS poster I present a number of allomorphic phenomena in Modern Hebrew and discuss whether learning them would require the learner to be sensitive to syntactic structure, phonological patterns, or both.
Itamar Kastner. Arabic consonants as bootstrapping for segmentation and acquisition. In the 10th Mediterranean Morphology Meeting (MMM10). Joint work with Frans Adriaans.
[info +/-]

Superseded by Kastner and Adriaans (2017).
Itamar Kastner. Nonactive voice in Hebrew and elsewhere: Between unaccusativity and agentivity. In the 39th annual Penn Linguistics Conference (PLC 39).
[info +/-] [paper]

Two verbal templates in Modern Hebrew allow for any kind of verbal construction, except for a simple transitive verb. Unaccusatives, reflexives, reciprocals and unergatives that take an obligatory indirect object are all attested, but transitive verbs are not allowed. I discuss what the morphology of these templates actually signals, given that external arguments and internal arguments are both possible. Working in Distributed Morphology, I propose that a number of functional heads conspire to produce the existing alternations in argument structure, with implications for theories of anticausativization, reflexivization and reciprocalization.
2014 Itamar Kastner. Nonconcatenative Morphology with Concatenative Syntax. In the 45th annual meeting of the North East Linguistic Society (NELS 45), Special Session on the Phonological Consequences of Morphological Structure.
[info +/-] [paper]

In this NELS talk I present a DM-based system that predicts what the argument structure of a given Hebrew root will be in a given verbal template.
Itamar Kastner and Vera Zu. Paradigm Gaps in the Hebrew Passive: Licensing, Referentiality and Agentivity. In the 32nd West Coast Conference on Formal Linguistics (WCCFL 32).
[info +/-] [poster] [abstract]

In this WCCFL poster, Vera Zu and I present a new puzzle in Hebrew morphology and try to tie it in to the crosslinguistic behavior of implicit grammatical elements.
2013 Itamar Kastner and Kate Davidson. Nominalizing Clauses: Evidence from ASL and a New Typology of Embedding. In Theoretical Issues in Sign Language Research (TISLR) 11.
[poster]
2012 Itamar Kastner. Cognate Objects and Unaccusative Predicates. In the 14th SUNY Stony Brook, Yale University, New York University and CUNY Graduate Center Conference (SYNC 2012).
[slides] [abstract]