Poster session  |  Session d'affiches

1 Introduction. Compounding is a process of deriving complex morphological forms by the merging of two or more uncategorized roots, such as hand-bag (Harley 2009). However, compounds with non-root units, such as German Kind- er-bett (‘child-LE-bed; crib’), and Icelandic vél-ar-hjól (‘machine-LE-wheel; motorcycle’) challenge this definition. Lieber & Štekauer (2009) describe non-root units in compounds as meaningless elements occurring between the two members of a compound, while Nübling & Szczepaniak (2009, 2010) argue that they are merely inserted as phonological repairs. Recent studies (Bauke 2014, Harðarson 2017, De Belder 2023) suggest that compounding is a syntactic process, treating non-root units as syntactic units. This study adopts a Distributed Morphology approach (Halle & Marantz 1993, 1994), and analyzes compounding as a cyclic and incremental process, supporting the view that morphological structures emerge from syntactic operations rather than being lexically pre-specified in the lexicon. Although non-root units are common in Germanic compounds, their functional roles in compounding remain an open question. Thus, the properties and distribution of non-root units in compounds, as well as the syntactic locality of compound constituents, may vary across related languages. This study investigates the role of non-root elements in Germanic compounds, their impact on syntactic locality and morphological exponency, and the distinction between Root Compounds (RCs) and Phrasal Compounds (PCs) within a cyclic domain.

2 Data. Compounds in German and Icelandic align with Nóbrega (2020, 2023) bipartite typology. Root Compounds (RCs) consist of two bare roots or a bare root and an nP head, while Phrasal compounds (PCs) comprise two nPs. The following examples illustrate RCs (1) and PCs (2) in both languages.

(1)  a. Hand-tasche                               b. hand-knatt-leikur             (2)  a. Tag-es-licht                        b.    karl-a-hesta-a-vagn
          hand-bag                                              hand-ball-game                             day-GEN-light                       men-GEN-horse-GEN-wagon
        ‘hand bag’ [German]                      ‘hand ball’  [Icelandic]                 ‘daylight’ [German]               ‘a horse carriage for men’[Icelandic]

The categorization of members of both types of compounds is linked to the presence of a functional head such as n, with syntactic derivation proceeding cyclically, where each morphological unit defines its own locality domain (Har ðarson, 2017). My treatment of this data makes the following predictions: (1) compounding occurs in cycles, with structural position determining locality constraints, (2) non-root units are syntactic exponents rather than phonological placeholders, and (3) grammatical function shapes compound structure.

3 Analysis. Following Distributed Morphology (Halle & Marantz 1993, 1994; Harley & Noyer 2009), I argue that compounds are derived cyclically, with each morphological unit forming a locality domain, becoming inaccessible after additional cyclic nodes are merged (Embick 2010). Historically, non-root units in German compounds evolved from case or plural markers. Building on the historical German and the synchronic Icelandic data, I treat non-root units that belong to the inflectional paradigm of compound modifiers, as cyclic heads i.e., little n, which trigger spell-out, forming an inner cyclic node where Root allomorphy is determined. The modifier’s cyclic node then merges with the compound head’s, creating an outer cyclic node that may block further interaction between constituents Furthermore, I argue that non-root units that do not belong to the inflectional paradigm of compound modifiers function as outer- cyclic nodes, preventing the root from determining allomorphy. This analysis highlights distinctions in the syntactic phases of compounding in Germanic languages.

4 Conclusion. An investigation of compounding in German and Icelandic using the Distributed Morphology approach supports the view that compound formation follows syntactic principles rather than separate lexical processes. This work offers empirical evidence for cyclic derivation in compounds and provides comparative perspective on Germanic compounding process.

“Mobile affixation” describes affixes that occur variably as prefixes, suffixes and infixes, determined by phonological or morphological features on the stem. Only relatively few cases of Mobile affixation have been discussed in the literature, including Polish (Embick, 1995), Huave, (Kim, 2010; Zukoff, 2021), Swahili and Fula (Stump, 1993), Moro (Jenks & Rose, 2015) and Afar (Fulmer, 1991).

In this paper, I investigate a small part (focussing on two of the ten verb classes) of the complex verbal paradigm of Koasati, which features mobile affixation. Some key data are in Table 1: note that the 2.sing, 1.pl and 2.pl are prefixes in ‘Class 1A’ verbs and infixes in ‘Class 1B’ verbs.

                                         Class 1A                                      Class 1B

Gloss                       ‘sleep’          ‘hear’                       ‘kindle’        ‘buy’

2.sing                       ís-nó ː ci        is-há ː lo                    ós-ti             cóspa

3                              nó ː ci            há ː lo                        ó ː ti              có ː pa

1.pl                          il-nó ː ci             il-há ː lo                    ól-ti              cól-pa

2.pl                          has-nó ː ci     has-há ː lo                o-hás-ti        co-hás-pa

I show that to analyse even a subset of this system requires a mixture of morphological and phonological approaches. Existing descriptions of Koasati use mainly morphological criteria (10 abstract ‘verb classes’) to index which roots take prefixes, suffixes and infixes. There are some generalisations associated with each class (e.g. ‘the roots that make up this class [1A] all have the syllable structure CVCV or iCV. However, not all verbs with this syllable pattern belong in this class’, Kimball, 1991:57), but, Kimball claims, class membership is not predictable from form.

I propose that a different analysis of the phonological shape of the roots (specifically, that class 1A forms have an underlyingly long and stressed initial vowel, whereas class 1B does not) allows us to claim that class membership is predictable from form in some cases, reducing the number of inflection classes needed. This allows us to account for correlations between stress and infixation, as well as reducing the amount of information that must be stored in the lexicon. However, there are some aspects of Koasati mobile affixation cannot be attributed to phonology (in particular, heteroclitic roots (Stump 2006), which ‘switch’ classes depending on their features). Therefore, Koasati requires a solution involving a mixture of phonological and morphological allomorphy.

This work and the data within it also offer ground to explore theoretical issues: in particular, the differences between approaches that suggest the placement of infixes must be phonologically optimising (e.g. Yu, 2003 and the ensuing discussion of ‘displacement theory’) and approaches that do not (eg. Kalin, 2022 and Condition Of Insertion (COIN)/Condition on Placement (COP)). The analysis here takes a dataset that is not immediately recognizably optimizing (which is why Kimball assumes that the different affix placements must be a morphological diacritic) and reanalyses it as optimizing. Obviously, a subset of data from one language is insufficient to make clear claims about either hypothesis: but it suggests that even data that does not appear to be optimizing may be analysed as such if the underlying forms are reconsidered; for this dataset, COIN/COP are more powerful than is necessary.

Prefixes, particles, and resultative secondary predicates (RSPs)  in German all have the ability to affect the argument structure of the complex predicate in which they appear. They are also almost entirely incompatible with each other, with the systematic exception of particle-prefix-verb combinations. I argue that the co-occurrence restrictions between prefixes, particles, and RSPs and their respective argument-structural properties are best understood as arising from a combination of phrase-structural considerations and a ban on multiple end states in a single event. I will argue that framing the analysis in terms of conditions on allosemy in the domain of complex predicates offers a way of understanding how the preverbal elements and the base verb come together. In particular, locality conditions on allosemy explain why RSPs seem to interact with the verb for argument introduction, but are always interpreted transparently.

The Welsh copula has positive (r-ydw ‘pos-be.prs.1sg’) and negative (d-ydw ‘neg-be.prs.1sg’) forms in the present and imperfect, as well as neutral forms (ydw ‘be.prs.1sg’). Sometimes, the neutral occurs where the positive or negative might be expected (given the polarity of the clause), such as in response to a present or imperfect polar interrogative. In Welsh, responses to polar interrogatives require the use of an auxiliary verb (there is no ‘yes’ or ‘no’), and only neutral ‘be’ is used. ‘Yes’ is ydw ‘be.prs.1sg’ and ‘no’ is nac ydw ‘neg be.prs.1sg’, not r-ydw ‘pos-be.prs.1sg’ and (nac) d-ydw ‘(neg) neg-be.prs.1sg’. The blocking of positive and negative forms is not symmetric, with the former being blocked in more contexts than the latter. 

A previous analysis within HPSG (Borsley 2019) ties the distribution to clause type, and treats the neutral copula as equally as specified as the positive and negative. This approach faces difficulties in if-clauses, where the positive is blocked only in ordinary cases, and not when focus-fronted material intervenes between ‘if’ and the copula. I develop an account within LRfFG (Melchin et al. 2020; Asudeh et al. 2021, i.a.). I propose that the neutral forms should be treated as the default, unmarked form, which is ordinarily blocked by the availability of more (polarity-)specific forms, but occurs when these forms are incompatible with the context. This incompatibility happens more often in the case of the positive, because the r- morpheme spans both the complementizer and polarity heads of the tree. These forms show us that even though positive polarity is the default polarity value, it is not an unmarked polarity value, and therefore polarity must be a binary rather than privative feature in LRFG. The data also sheds light on the nature of blocking in LRFG, particularly concerning the role of negative constraints, and the prioritization of spans where possible.

This presentation examines how the nature of morphosyntactic features, particularly phi features, interacts with XP movement of the verbal stem (TP) to explain discontinuous agreement in Semitic languages.

This exploration may provide a new perspective on the long-standing debate concerning the Doubly Filled Comp (DFC) filter. The DFC filter, proposed by Chomsky and Lasnik (1977), was generalized by Koopman (1997), in terms of projection activation by postulating a modified version of Kayne's (1994) LCA (Linear Correspondence Axiom). A projection is activated by open lexical material, either in its head or in its specifier, but never in both positions simultaneously. However, although there are complementarity effects in the DFC filter, there are plausibly cases where both positions are realized openly. It therefore appears necessary to develop a method for deriving the effects of the DFC filter that is both general and parameterizable.  

In this study, I adopt Koopman's projection activation principle, but without modifying Kayne's LCA. I propose an alternative means of deriving the effects of the DFC filter. Adopting the granular mapping (OFOH) of Cinque and Rizzi (2008), I argue that the two projection activation options, head lexicalization (lex) and specifier lexicalization (EPP), are properties common to all features. However, I propose that the way this manifests itself depends on the nature of the features.

For binary features, [±f], each value has one of the lexicalization options (lex) head lexicalization, or (EPP) specifier lexicalization. For privative features [f], on the other hand, both properties (lex and EPP) are simultaneously active, meaning that they require lexical material both in the head (lex) and in the specifier (EPP). This proposition will be tested and illustrated through the syntactic analysis of discontinuous verb agreement in Standard Arabic.

For decades, there has been debate surrounding how gender is assigned to simple English loanwords in Spanish (e.g., selfie) because, unlike English, Spanish nominals require a grammatical gender assignment to a particular category (either masculine or feminine). Two major hypotheses argue that either the terminal phoneme/grapheme (Bull, 1965; Bergen, 1978) or the default masculine (Harris, 1991) dictate grammatical gender assignment in Spanish. Another hypothesis that exists, but that is underrepresented, argues in favor of analogical gender (Sánchez, 1992) dictating the gender assignment. These hypotheses have mainly accounted for simple loanwords, but not for compound loanwords specifically (e.g., fashion week). This study adopts a(n) Distributed Morphology approach/exoskeletal model of grammar (e.g., Borer 2005a, 2005b; Halle & Marantz, 1993; Grimstad et al., 2018), arguing in favor of analogical gender dictating the gender assignment of English compound loanwords in Spanish. Specifically, this study analyzes 75 loanwords from Twitter/X (and subsequent speaker elicitations). The dataset argues for gender assignment via a semantic association with the Spanish equivalent. The first class’s gender assignment, which is more common statistically, is dictated by the Spanish equivalent of the compound’s second element (el reality show <<< el programa). The second class’s gender assignment is dictated by the Spanish equivalent of the compound’s first element (la reality show <<< la realidad). I argue that these compound loanwords are generated in separate workspaces and enter the structure as heads (Harðarson, 2017), which deviates from previous accounts (Harley, 2009). Gender is located on n (Kramer, 2015). 

I explore the cross-linguistically long-attested dichotomy between 1st and 2nd person on the one hand,

and 3rd on the other, in the case of Turkish verbal inflection. Motivated by both syntactic and morphological data, I

decompose pronouns and verbal inflectional morphology and claim that 3rd person in Turkish does not have person

features the way that 1st and 2nd person do. I propose a syntactic analysis for person verbal inflection: a clitic-doubling

account in the case of 1st and 2nd person, and in the case of third person, that what looks like person/number agree-

ment is really zero-number agreement coupled with cliticization. Along with this, I discuss the morphological decom-

position of four different person inflection categories, giving the vocabulary items and insertion rules underlying the

syncretisms across the two nominal and two verbal categories within the framework of Distributed Morphology (Halle

and Marantz, 1993; Embick, 2015). Combining cartographic and late insertion perspectives, this work suggests a fairly 

novel larger approach to these morphosyntactic issues.

Basque finite auxiliaries mark the person and number of their ergative, absolutive, and dative arguments. The prevailing view is that at least some of these markers are clitics, though there is some disagreement about exactly which ones. I present data like (1) which shows that Markina Basque permits summative agreement, creating a problem for clitic-doubling analyses.

(1)        [Jon-ek   ure]                eta   [ni-k       teie]           edaten        do-gu-z.

   Jon-ERG               water.ABS      and  1SG-ERG  tea.ABS   drink.IMPF    AUX-1PL.E-PL.A

   'I drink tea and Jon water.'

Here the auxiliary seems to get the 1PL ergative marker -gu by "combining" (in some sense) the ergative arguments from two separate clauses, and the same thing seems to happen with the absolutive arguments. However, if we are inclined to view -gu as a 1PL clitic, there is no 1PL DP in which it could originate. The same problem can be demonstrated for absolutive and dative arguments. Building on a framework developed by Grosz (2015), I show that this problem can be solved under an agreement-based analysis. I also discuss several possible ways to repair the clitic-doubling analysis and show that they either fail to explain the data or lead to larger theoretical problems.