Program

Mixed languages are a disputed category of contact languages defined as languages whose essential components derive from two (or more) languages, making their genetic classification problematic (e.g., Meakins 2013; Bakker 2017; Versteegh 2017; Hölzl 2024). For example, Media Lengua combines the grammar of Quechua with a lexicon from Spanish. While the concept is often (implicitly) understood as involving genetically unrelated languages, this talk distinguishes between polyphyletic mixed languages, which combine elements of unrelated languages, and monophyletic mixed languages, which combine elements from different branches of the same language family and thus resist classification at the level of individual branches. The latter concept is illustrated with Kilen, a Tungusic language that combines elements associated with all four major branches of Tungusic and has consequently been assigned different positions within the family (e.g., Kazama 1998). Kilen thus demonstrates that language mixing can pose a challenge not only to the traditional classification into language families, but also to the subclassification of languages within a single family (e.g., Hölzl 2025).

Bakker, Peter. 2017. Typology of mixed languages. In Alexandra Y. Aikhenvald & R. M. W. (eds.), The Cambridge handbook of linguistic typology, 217-253. Cambridge: Cambridge University Press.

Hölzl, Andreas. 2024. Mischsprache [mixed language]. In Mechthild Habermann & Ilse Wischer (eds.), Historische Sprachwissenschaft (Wörterbücher zur Sprach- und Kommunikationswissenschaft (WSK) Online). Berlin: Mouton de Gruyter.

Hölzl, Andreas. 2025. Studies in East Asian Comparative Linguistics. Potsdam: University of Potsdam. (Cumulative habilitation thesis.)

Kazama, Shinjirō. 1998. Genetical position of Hezhen. In Osahito Miyaoka & Fubito Endo (eds.), Languages of the the North Pacific Rim, Vol. 9. Kyoto: Nakanishi.

Meakins, Felicity. 2013. Mixed languages. In Peter Bakker & Yaron Matras (eds.), Contact languages. A comprehensive guide, 159–228. Berlin: De Gruyter Mouton.

Versteegh, Kees. 2017. The myth of the mixed languages. In Benjamin Saade & Mauro Tosco (eds.), Advances in Maltese linguistics, 217–238. Berlin: Mouton de Gruyter.

It has long been assumed that articleless languages compensate for the lack of grammatical definiteness by flexible word order and/or information structuring (Krámský 1972, Birkenmaier 1979, Geist 2010, a.o.): clause-initial (or preverbal) bare (= determiner-free) nominals correspond to definites and clause-final (or postverbal) bare nominals correspond to indefinites. (Cf. also the relative-order approach of Titov 2012.). In this talk, I report on a number of recent studies addressing the issue in an empirically rigorous way (Šimík & Burianová 2020, Śimík & Demian 2020, 2021, Nováková 2026, Jeřábková & Šimík in prep). The studies give a non-trivial picture: while corpus studies confirm the hypothesized relation between word order and determinacy (revealing a correlation between clause-initiality vs. finality and referential determinacy vs. indeterminacy), experimental studies fail to support the hypothesis that the semantics of definiteness corresponds to the semantics of clause-initiality. We are thus left with an interesting open question: which semantic or pragmatic notion does word order encode?

Grammatical categories such as tense, mood, person, and voice are often expressed by combinations of words rather than by individual word forms. While morphological annotation typically assigns grammatical features to individual words, the properties of analytic constructions as a whole also need to be represented to facilitate their identification and comparison across languages.

We present a rule-based approach to identifying analytic verb forms and assigning morphological features to entire verbal constructions, using existing morphological and syntactic annotation in Universal Dependencies (UD). Implemented in the Udapi framework, the tool covers most Slavic and Romance languages. We illustrate the approach with examples from both language groups, discuss differences in the expression of grammatical categories, and present evaluation results for Czech and Ukrainian.

We also show how related information is represented in InterCorp, the multilingual parallel corpus, through content-word features: an additional annotation layer that makes grammatical information contributed by dependent function words accessible at the level of their content-word head. Examples of corpus queries demonstrate how this annotation facilitates the retrieval and crosslinguistic comparison of grammatical constructions.

Finally, we discuss some broader challenges of multilingual morphological annotation, illustrated by recent changes in the French UD treebank, and introduce the latest release of InterCorp, including new annotation features and other developments relevant to contrastive linguistic research.

Selice Romani, a variety of Romani spoken by Romani–Hungarian bilinguals in a village in southwestern Slovakia and strongly influenced by Hungarian (e.g. Elšík 2007, 2009, 2022), has several oikonyms (i.e. names of towns and villages) that are lexical borrowings of Hungarian inessive–illative case forms, e.g. H Diáki‑ba [D‑ine/ill] ‘in/to Diakovce’ (← Diáki ‘Diakovce’), but have been reanalysed as nominatives, e.g. R Diákib‑a [D(f)‑sg.nom] ‘Diakovce’, serving as bases for further inflectional forms, including spatial case forms, e.g. R Diákib‑a‑te [D(f)‑sg.obl‑loc] ‘in/to Diakovce’. In some cases the reanalysed oikonyms are morphologically adapted through mere resegmentation, i.e. through the reanalysis of the final /a/ of the Hungarian inessive–illative form as a singular nominative marker in Romani, as in the above example, while in others they are adapted by an overt suffix, e.g. R Selib‑a ‘Saliby’ < H Szeli‑be ‘in/to Saliby’ (← Szeli ‘Saliby’). In yet another instance, a resegmented reanalysed oikonym has diverged from its source form through a subsequent sound change: R Požom‑a < *Požomb‑a ‘Bratislava’ < H Pozsom‑ba ‘in/to Bratislava’ (← Pozsony ‘Bratislava’) (Elšík 2017). In our presentation of the data, we will draw attention to the methodological challenge arising from the small dataset of reanalysed oikonyms.

In the case of the resegmented (and phonologically unchanged) reanalysed oikonyms the Romani–Hungarian bilinguals use phonologically identical forms in both languages (i.e. in repertoirs or sets of structures used in, and pragmatically indexed for, different social and communicative contexts, e.g. Matras 2009, 2025), which however have different grammatical functions. This can be described in terms of the useful distinction between pro-diasystematic and counter-diasystematic linguistic change (e.g. Höder 2018 and Boas & Höder 2025). On the one hand, the material borrowing of the Hungarian oikonyms by itself represents a pro-diasystematic innovation, i.e. a type of innovation that decreases the context-specificity of linguistic structures in multilingual settings. On the other hand, the functional reanalysis (and the morphological adaptation) of the borrowed oikonyms shows that the change also has counter-diasystematic (or, equivalently, pro-idiosystematic) aspects, in that the materially diasystematic structures have developed context-specific, idiosystematic, grammatical functions.

In this talk, we will contextualise the reanalysed oikonyms in Selice Romani in several ways. First, we will briefly outline the multilingual situation of the Selice Romani speech community. Second, we will describe Selice Romani lexical borrowings of Hungarian case-marked forms (of various nominals) that have not been functionally reanalysed. Third, we will contrast the reanalysed oikonyms with Selice Romani oikonyms that are lexical borrowings of Hungarian nominative (rather than inessive–illative) forms. Finally, we will present examples of reanalysed oikonyms from other languages and contact situations as typological parallels. The talk will conclude with several preliminary theoretical observations. First, we will discuss the potential motivations for the reanalysis, such as textual frequency of the source forms and their diasystematic potential. Second, we will give some thought to the idea of degrees of pro-diasystematicity in language change, discussing concepts such as fusion or non-separation of languages with respect to an entire functional category (e.g. Matras 1998, 2000), and to the phenomenon of counter-diasystematic aspects of borrowing in the context of the Diasystematic Construction Grammar framework (e.g. Höder 2018 and Boas & Höder 2025). Finally, we will briefly outline the prospects for integrating our observations into a framework of cognitively realistic description of linguistic change in multilingual contexts.

[Abbreviations: H = Hungarian, ine/ill = inessive-illative, loc = locative case, nom = nominative case, obl = oblique stem, R = Selice Romani, sg = singular number]

Czech pronominal clitics are typically described as occupying a position after the first syntactic constituent of a clause and obeying a fixed templatic order of reflexive–dative–accusative. In this talk, I present the results of two corpus studies that complicate this description. First, I look at the cooccurrence of accusative and genitive clitics, a rare combination described inconsistently in previous work. The order of these clitics is variable and depends on the clause’s argument structure, unlike the much more common combination of dative and accusative clitics. Second, I look at the factors influencing whether clitic arguments of infinitives appear in the infinitival clause or in the matrix clause (known as „clitic climbing“). In this case, the availability of the two clitic positions depends heavily on the matrix verb. I discuss the implications of these studies for proposals in generative syntax that locate variation in the variable choice between two mutually incompatible categorical grammars, rather than within a single variable grammar.

The talk will provide insight into an emerging master thesis whose aim is to investigate the functions of the indefinite pronouns in the Chuvash language. The presentation will focus on the preliminary research, theoretical anchors, and specifics of working with a less-described language. Chuvash is the only surviving member of the Oghuric (Bulgharic) branch of the Turkic family. Being the first one to have branched off the family tree, it is considered its most divergent member (Savelyev, 2020). The indefinite pronoun systems in more widely spoken Turkic languages, let alone Chuvash, remain understudied. According to reference works, Chuvash features three series of interrogative-based indefinite pronouns: those with the prenominal tA- particle (takam ‘someone’), those with both the postnominal tA- and the pulin or pulsan particles (kam ta pulin ‘anyone, whoever it may be’), and those with the prenominal kirek particle (kirek kam ‘somebody’). Other reference work also mentions the postnominal tĂr particle (kam-tăr ‘someone, somebody’). This system is in striking contrast with the remainder of the family, which aligns with the degree of divergence mentioned above. Methodologically, the thesis will follow Haspelmath’s framework of semantic maps. Apart from theoretical modelling, the study will involve data elicitation and acceptability judgements from the speakers of the language following the methodology of semantic fieldwork proposed by Matthewson (2004), as well as available language corpora as an auxiliary means. In the second part, the study will focus on peculiarities of the Chuvash indefinite pronouns and propose their formal semantic description using a formal apparatus inspired by works from related domains.

This talk explores key grammatical features of Mawes, a language isolate spoken along the northern coast of West Papua, Indonesia, in a highly diverse region characterized by extensive contact between coastal and mountain communities. I examine its four-way number system (singular, dual, paucal, plural), hodiernal suffixes specifying time of day, and object marking. I also discuss the language’s sociolinguistic situation, including the threat posed by Indonesian, and introduce ongoing community-based documentation and revitalization efforts. Finally, I consider how intermarriage and trade with neighboring groups have shaped the Mawes lexicon, particularly in the domains of colour, the sea, technology, and numerals. This case study contributes to our understanding of Papuan typology, highlights under-documented grammatical features, and illustrates broader processes of language contact and shift.

Questions are a frequent form of replies expressing surprise. Using examples of literary dialogue in Polish, Russian and English, I discuss how interrogative form relates to the expression of surprise and to other functions of a reply. The examples include echo questions, requests for confirmation and questions seeking information or an explanation. Particular attention is given to short interrogative replies whose interpretation depends on the preceding turn, and to longer replies in which a further question makes the source of surprise more explicit. Drawing on Awdiejew’s functional approach to pragmatics, I distinguish the matter addressed by the question from the speaker’s emotional response to it. The analysis considers narrative descriptions of emotion alongside the interlocutor’s subsequent response, while recognising the limitations of literary dialogue as evidence of spoken interaction. I argue that the expression of surprise can accompany a genuine request for information and discuss what this coexistence implies for the interpretation of indirectness.

This talk presents some of the results of joint work with Karen De Clercq, Michal Starke and Guido Vanden Wyngaerd on various aspects of Czech degree morphology (Caha et al. 2019, 2024, 2026, Vanden Wyngaerd et al. 2024).

Our (ever continuing) investigation is inspired by Jonathan Bobaljik’s work on the decomposition of adjectival degrees. In his 2012 book, he surveyed a sample of over 300 languages and showed, based on a variety of evidence, that the positive, comparative and superlative degrees are decomposable and stand in a containment relation: the superlative is built on top of the comparative, which in turn is built on the positive degree adjective. Among the evidence adduced for this conclusion was the observation that the superlative degree often morphologically contains the comparative as a subpart (as it does in Czech), and that the comparative in turn typically morphologically contains the positive (ignoring agreement). In some approaches, the containment of the positive inside the comparative has been posited as a candidate universal.

Against this background,  Czech provides a much more nuanced picture regarding the relationship of the positive and the comparative: we find cases where the comparative is derived from the positive by the addition of two or even three independent suffixes, while in other cases the comparative is not derived from the positive, but from a shorter, truncated form of the positive. These intricate patterns of morphological derivation between the positive and the comparative do not at first seem to fit the general pattern and invite a further investigation. 

In the talk, I argue that the full morphological complexity can be derived by unifying two strands of research: on the one hand the original claim that the comparative syntactically contains the positive; and on the other hand decomposing the adjective into more fine grained ingredients familiar from the semantics research distinguishing scales, standards of comparison, measurement, etc.

This talk investigates how different assumptions about the representation of person and number constrain the space of possible mixed perfective auxiliary paradigms in Romance. The empirical focus is on Italo-Romance compound tenses systems in which auxiliaries descending from ESSE (E) and HABERE (H) alternate within a single person–number paradigm, as documented in the MIXPAR database. I begin with a more general question of feature representation, distinguishing between feature ontology, understood as the inventory and nature of representational primitives, and feature system, understood as their concrete mapping onto paradigm cells. This distinction matters because different analyses of person and number make different natural classes available. In ongoing collaborative work, I examine how alternative feature systems affect the descriptive complexity and naturalness of mixed auxiliary paradigms. The second part asks what happens when these feature distinctions are given an explicit syntactic organisation. I compare two independently motivated nanosyntactic orderings, number below person and person below number, within a deliberately restricted containment model. Both generate 14 of the 64 logically possible categorical E/H patterns, but they select substantially different subsets. Crucially, only the person-below-number ordering directly derives the dominant MIXPAR pattern EEHEEH, in which the same participant split recurs across singular and plural. More generally, mixed auxiliation is used here as a test case for the relation between feature ontology, morphological natural classes, and syntactic containment.

How do speakers derive meaning from unfamiliar words? A growing body of research suggests that word processing relies not only on stored lexical representations but also on sublexical information, which can guide the interpretation of rare words and novel letter strings. But what are these sublexical units, and how do they relate to morphology?
In this talk, I will explore how distributional information associated with recurring letter sequences contributes to semantic processing, and how this information relates to morphology. Drawing on behavioural data, computational modelling, and neuroimaging evidence, I will examine how speakers process rare unfamiliar words and pseudowords, and how semantic information can be activated by both identifiable morphemes and non-morphological sequences of letters. Across studies, a consistent picture emerges: morphology contributes to semantic processing, but it does not exhaust the sources of meaning available at the sublexical level. Character sequences that do not correspond to morphemes can activate semantic information and influence behaviour, while models incorporating explicit morphological structure can offer additional explanatory power.
I will argue that morphological units can be understood as particularly stable and systematic instances of form–meaning mapping within a broader landscape of sublexical regularities. This perspective raises further questions about the nature of sublexical representations, the relationship between morphological structure and distributional learning, and the ways in which computational models capture the information that speakers exploit when processing unfamiliar forms.