The schedule and recordings are posted below.
You can access all slides and handouts by clicking here.
|Friday December 2|
|8:45 – 9:35||— Breakfast —|
|9:35 – 9:45||— PɸF Welcome —|
|9:45 – 10:30||Peter Arkadiev & Yury Lander [virtual]||Ambifixes: cross-linguistic variation and possible origins [recording]|
|10:35 – 11:20||Milena Šereikaitė||The dual nature of Lithuanian reflexive -si- [recording]|
|11:20 – 11:50||— Coffee Break —|
|11:50 – 12:35||Nicholas Rolle [virtual]||The morphological/phonological behavior of bipartite morphemes [recording]|
|12:35 – 1:45||— Lunch Break —|
|1:45 – 2:30||Nikita Bezrukov||Mobile verb markers in Erzurum Armenian and the theory of word formation [recording]|
|2:35 – 3:20||Hossep Dolatian||Cyclic residues of affix deletion in Armenian passives [recording]|
|3:20 – 3:50||— Coffee Break —|
|3:50 – 4:35||Stela Manova||The wheres and whens of affixation: theoretical, computational, and experimental evidence [recording]|
|Saturday December 3|
|8:45 – 9:45||— Breakfast —|
|9:45 – 10:30||Johanna Benz||French verbs: decomposition and the rules of morpho-phonology [recording]|
|10:35 – 11:20||Yuni Kim [virtual]||Derivational ordering in Amuzgo non-linear exponence [recording]|
|11:20 – 11:50||— Coffee Break —|
|11:50 – 12:35||Sam Zukoff||Structure and base-derivative correspondence in Bantu affix ordering [recording]|
|12:35 – 1:45||— Lunch Break —|
|1:45 – 2:30||Alan Yu||The expressive origins of expletive infixes [recording]|
|2:35 – 3:20||Laura Kalin||How many ways are there to move? The view from infixation and beyond [recording]|
|3:20 – 3:50||— Coffee Break —|
|3:50 – 5:00||Mary Paster [virtual] & Karlos Arregi||Discussion session [recording]|
|5:15 – 7:15||— Reception —|
Peter Arkadiev & Yury Lander
“Ambifixes: cross-linguistic variation and possible origins”
We shall start by providing a working definition of ambifixes as affixes that can occur both as prefixes and suffixes. Then we shall outline a typology of ambifixation constructions based on the types of conditioning factors determining the position of the ambifix. Basing ourselves on a cross-linguistic convenience sample of ca. 40 languages, we shall discuss and exemplify ambifixes subject to phonotactic, morphotactic, morphosyntactic, lexical, and semantico-syntactic factors. Besides that, despite the necessarily limited evidence, we shall try to look into the possible historical origins of ambifixation, singling out such diachronic paths to ambifixation as affixalisation of variable-position clitics and univerbation of constructions containing auxiliary elements bearing an affix in the “right” position by as a whole occuring “on the other side” of the lexical element.
“French verbs: decomposition and the rules of morpho-phonology”
Morpho-phonological rules are perhaps the least-loved component of Distributed Morphology, and have drawn criticism over the last thirty years which repeats three main concerns: That morpho-phonological rules are a technical trick to fix something the theory has nothing interesting to say about; that such rules are too powerful, in principle capable of rewriting any string as any other string; and, finally, that they are non-modular, violating a strict division between syntax and phonology. Within and outside of DM, this has led to the conclusion that many phenomena that involve morphologically-conditioned alternation should be handled as (weak) suppletion. Based on a case study of French verb root alternations, I offer a series of arguments for the position that morpho-phonological rules are an insightful analysis of this putative `weak suppletion’ phenomenon, showcasing in particular the importance of considering the specific rules in the context of both the morphological decomposition and regular phonology of the language in question.
“Mobile verb markers in Erzurum Armenian and the theory of word formation”
This talk zooms in on the cross-dialectal distribution of the Indicative/Conditional marker in Modern Armenian. The marker in question, which I label as K, is realized as a prefix in the majority of Armenian varieties. Nevertheless, in a subset of Western(ish) varieties, it has developed a mobile distribution within a complex head, switching between a prefix and a suffix. Erzurum Armenian exhibits a diachronic tension between allowing the K marker to have a clitic-like clausal distribution and forcing it to affix to the verb, which lead to a variety of outcomes across the subvarieties/speakers. I document the variation and provide a structural account of it. This case study is theoretically informative as is shows that mobile affixation on a head can be affected by clausal syntax. This is an important piece of evidence against morphological models which keep clause building and word building as separate non-interacting components of the grammar. Furthermore, Erzurum Armenian shows that selecting a head for pronunciation in a sequence of heads related by head movement is separate from head movement itself since some speakers choose low pronunciation while others prefer high or mixed pronunciation.
“Cyclic residues of affix deletion in Armenian passives”
Morphophonological derivations are often simultaneously isomorphic to both surface morphotactics, cyclic phonology, and to abstract morphosemantic structure. On the surface, passive verbs in Armenian violate this isomorphism. Verbs consist of a bound stem plus suffixes. Although passive verbs seem morphologically derived from roots, their phonological structure is only predictable from the pre-vocalic stem of active verbs. I capture this predictability by combining cyclicity with affix truncation. The interaction between cyclicity and affix deletion causes the opaque overapplication of phonological rules in passives, because such rules applied cyclically early in the active verb. This approach is a cyclic alternative to using output-output constraints. However, our cyclic analysis likewise predicts *ABA effects in the morphosemantics of passives. Thus, our analysis creates a cyclic dependency between two bound stems that is visible across modules.
“How many ways are there to move? The view from infixation and beyond”
Across the literature, it’s easy to find well over a dozen different kinds of apparent “movement” in some component of the grammar (from phonology through syntax). This talk begins to explore the following question: How many distinct ways are there really for grammatical elements to “move”? Answering this question involves (i) developing diagnostics for empirically distinguishing among—or unifying, as the case may be—the wide variety of phenomena that seem to involve displacement of some sort, and subsequently (ii) balancing the need for empirical coverage with the need for a constrained theory that captures the cross-linguistic typology of (im)possible “movements”.
In this talk, I begin with the view from infixation: What kind of movement (if any) is involved in infixation, and what diagnostics can we leverage to distinguish it? I argue that infixation involves a specific kind of phonological operation that happens alongside exponence, involving displacement from a stem-edge position (where the morpheme is linearized) to a stem-internal one (where the infixal exponent is linearized). In diagnosing this behavior, I make crucial use of typological findings about infix allomorphy (Kalin To appear). In the second part of the talk, I catalog a number of distinct phenomena that seem to implicate movement, and the mechanisms that have been invoked to account for them. Finally, I suggest a path forward in terms of diagnostics and building a constrained theory of “movement” across the grammar.
“Derivational ordering in Amuzgo non-linear exponence”
Non-concatenative morphology, where grammatical features are exponed via phonological changes to a stem rather than by linear attachment of affixes, is not always obviously distinguishable from suppletive allomorphy. One common criterion for distinguishing the two is regularity and/or productivity, with irregular stem changes generally necessitating lexical listing and hence a suppletion analysis. Another criterion is the extent to which the stem changes can be described in terms of legitimate phonological units and operations; if they cannot, as in the case of English go and went, we have suppletion.
A problem with these criteria is that, while they can rule out a phonological analysis of stem changes, they can arguably never rule out suppletion. Sets of stem allomorphs that are regular and phonologically transparent are usually assumed to be derived from a single underlying form, ostensibly on the idea that what can be acquired as phonology, must be acquired as phonology.
In this talk, I examine some of the extensive stem alternations that characterize verbal paradigms in the Amuzgo (Oto-Manguean) language as spoken in San Pedro Amuzgos, Oaxaca State, in southern Mexico. In Amuzgo, I argue that we can distinguish suppletive allomorphy from subsegmental affixation by looking at the behavior of inflectional exponents in multimorphemic and phrasal verbs. When comparing various configurations of derivational morphology, we see that the structural ordering of derivation before inflection creates the potential to ‘disrupt’ inflection in ways that play out differently depending on whether stem allomorphs are selected based on morphological context, versus created via morphophonological operations. I analyze person features in terms of subsegmental affixation: the exponents surface in phonologically defined rather than semantically morpheme-bound locations, and their interaction with lexical inflectional classes is subject to locality conditions. Meanwhile, I develop an analysis of plural stems as suppletive allomorphs, despite their superficial compatibility with a morphophonological description. Suppletion correctly predicts observed patterns of multiple exponence, as well as the lack of locality conditions on the visibility of morpholexical information to plural marking.
The upshot is that while the grammatical distribution of some ‘phonological’-looking stem changes neatly fulfill the predictions of a serial, piece-based model of word formation, others may not. In the latter cases, such frameworks force us to conclude that the alternations are not created by the language’s phonological grammar, in apparent contravention of the popular principle that predictable information is the province of grammar, whereas only unpredictable information is stored in the lexicon. I therefore make a suggestion that is rather unoriginal from certain points of view, but perhaps controversial in a generative context: that models of the lexicon should allow for statements of phonological relationships between allomorphs which fall outside the phonological grammar proper, but which nevertheless form part of speakers’ competence. This purely lexical domain presents a novel possibility for the wheres and whens of “affix” ordering, in cases like Amuzgo where the standard menu of wheres and whens may or may not be adequate once the complexities of the derivational picture are considered.
“The wheres and whens of affixation: theoretical, computational, and experimental evidence”
I will claim that the answers to the questions about the wheres and whens of affixation depend on the approach followed.
With respect to the assumed relation between meaning and form, theories of morphology are of two types (Manova et al. 2020):
A) Theories that see form and meaning as inseparable: such theories operate with traditional morphemes, i.e. the morpheme is a minimal unit of structure that associates form and meaning (e.g., Minimalist Morphology and Natural Morphology);
B) Theories that separate meaning from form and are meaning-first (e.g., Distributed Morphology and Paradigm Function Morphology; the former operates with abstract morphemes, the latter is amorphous and works with paradigms).
By contrast, computational approaches to morphology (Hammarström & Borin 2011) are form-first (recent NLP research even generates language based exclusively on form, e.g., the n-gram model and the famous neural network BERT, Devlin et al. 2019).
Depending on the experiment design, experimental approaches are, as a rule, either form-first (e.g., a decision task involving visual recognition of existing and non-existing forms) or meaning-first (e.g., using pictures as stimuli for grammatical task solving).
Based on the above facts, I will present and discuss the results of a psycholinguistic experiment on the processing of affix ordering in various languages whose affix order systems were analyzed with the help of a mathematical method, the Gauss-Jordan elimination. The experiment was replicated with native and advanced non-native speakers of English and German, as well as with native speakers of Italian, Polish, Slovene, and Spanish.
I will support the surprising (from a theoretical point of view) results of the psycholinguistic experiment with examples of allomorph selection that are phonology-first (i.e. form-first) and involve information outside the word in which the selected allomorph occurs. My data will come from Slavic (from Bulgarian, but also from other Slavic languages, if the time allows).
“The morphological/phonological behavior of bipartite morphemes”
This paper investigates the behavior of ‘bipartite morphemes’ (Harris 2017: 17; Kuryłowicz 1966 [1945-1949]), where a single linguistic category is expressed by discontinuous parts which do not individually express separate categories. Common types include circumfixation (e.g. in (1), German participle morphology ge-…-t), and other discontinuous sequences such as West African ‘splitting verbs’, inter alia.
(1) Circumfixation, e.g. in German (Zingler 2022:60, i.a.)
‘(to have/be) googled’
We can initially formalize bipartite morphemes as a single morphosyntactic feature set [F] exponed as discontinuous phonological parts /α/ and /β/, in (2).
(2) [F] ↔ /α … β/
The central thesis of this talk is that these two parts /α/ and /β/ always behave independently from one another with respect to morphological and phonological behavior, i.e. they never act as a single morphological unit. This is exemplified in three ways – with respect to insertion, allomorphy, and derivedness – summarized in (3).
(3) Morphological independence of bipartite morpheme components
- Insertion: The conditions governing the (non-)insertion of one component never affects the other component
- Allomorphy: Allomorphy triggered by or targeting one of the components never affects the other component
- Derivedness: When the two components are incidentally adjacent, they always act as a derived environment with respect to morpho-phonological processes
Based on this thesis, we adopt that the individual components of bipartite morphemes constitute multiple ‘morphs’, defined as the minimal non-decomposable forms of a language (Haspelmath 2020). We couple this with a modified version of van Oostendorp’s (2005, 2007) ‘Morphological Color’ theory, where all phonological material of a single morphological unit are jointly indexed (i.e. have the same ‘color’). We conclude that a single feature bundle may be expressed by multiple discontinuous morphs (with separate colors, 4a), but individual morphs themselves are never discontinuous (4b). In short, individual morphs can be never contain ‘territorial exclaves’ in their underlying representation.
|(4)||a.||ok||[F] ↔ /αi … βj/||two morphs|
|b.||*||[F] ↔ /αi … βi/||*one discontinuous morph|
We conclude this talk by examining the implications of this thesis for understanding the fine typology of grammatical tone patterns.
“The dual nature of Lithuanian reflexive -si-”
(Abstract to come)
“The expressive origins of expletive infixes”
Expletive infixes (e.g., fan-bloody-tastic in English or pre-gwai-sent ‘present + gwai’ in Cantonese) are unique among infixes typologically not only because the morpheme being infixed is a free morpheme, but their distribution tends to be more variable than grammatical infixes (e.g., un-bloody-believable vs. unbe-bloody-lievable vs. bloody unbelievable). These properties have led some to argue that they are not infixes at all. In this talk, I argue that expletive infixes are formally no different from grammatical infixes. Their unique distribution can be traced to their expressive origins. A case study of expletive infixation in Cantonese is presented to support this argument.
“Structure and Base-Derivative Correspondence in Bantu Affix Ordering”
Bantu’s well-known “CARP template” (Hyman & Mchombo 1992, Hyman 2003) interacts with various other aspects of the languages’ phonology and morphology. This talk will show how an approach to CARP based on the “Mirror Alignment Principle” (Zukoff 2022) and bigram morphotactic constraints (Ryan 2010) interacts with morphosyntactic structure to help explain two problems related to morpheme ordering in Bantu: suffix doubling in Chichewa (Hyman 2003), and overapplication of spirantization in Nyakusa (Persohn 2017). It argues that a parallel approach driven by Base-Derivative Correspondence (Benua 1997) provides a unified and more parsimonious solution to these problems than do serial/derivational approaches.