Collaboration with Ebrar Besinci and James Griffiths (both University of Tübingen).

Clarification questions are special in that they do not ask for information on the initiative of the speaker. Rather, they are a particular type of response to an utterance by an interlocutor: when the speaker needs clarification about something the interlocutor said, be it out of disbelief, inability to place something or mishearing. As a kind of repair device they are crucial for making conversation work and establishing mutual understanding. Their special function seems to come with a special form in many languages: they often allow for marked prosody and fragmentary morphosyntax that would be inacceptable in other sentence types. We investigate experimentally how their special function affects their form across languages with quite different prosodic and syntactic configurations. Starting from October 2025.  

Collaboration with Klaus von Heusinger (University of Cologne) and Jet Hoek (Radboud University Nijmegen). 

The Right Frontier Constraint from discourse structure captures the intuition that the current discourse segment only attaches locally: either to the most recent previous segment or to one that is superordinate to the most recent. We can see this in pronoun resolution. In a short discourse like "Liz ran the best time. Alice was second. She was very happy.", we understand that "she" most likely refers to Alice, even though Liz has just as much reason to be happy, if not more.  That's because by the time we get to the pronoun "she", the sentence with Liz is not at the right frontier anymore. But what if there are several candidates available at the right frontier? This is where formal properties  come into play. We pursue two different hypotheses here: 1) The more syntactically and prosodically integrated the most recent clause is, the more available is the content of the previous clause for attachment. 2) The prominence of the proposition gets inherited by the individual referents that are contained in them. We tested the effect of syntactic subordination, connectives, prosody and typographical markers on propositional and individual anaphora across different discourse relations in a series of experiments. Find some results here. Ongoing since September 2021. 

Collaboration with Ana Aguilar Guevara (Universidad Nacional Autónoma de México), Rafael Herrera (El Colegio de México), Klaus von Heusinger (University of Cologne), Carolina Oggiani (Universidad de la República) and Julia Pozas Loyo (El Colegio de México). 

Proper names like "Akiko" or "Balthasar" work differently from common nouns like "tree" or "primaballerina": Proper names seem to refer directly to a unique individual with that name. Common nouns, in contrast, provide a description which helps to restrict the set of potential referents to those the description fits. That is why they combine with further modification and crucially, with determiners like definite or indefinite articles to actually determine their reference in a context. When I say, "I saw the marine biologist walking up to my house", the function of the definite article the is to pick out the uniquely identifiable and familiar individual from the context who fits the description marine biologist. Proper names, since they establish reference directly, don't need this additional determination: in "I saw Lucia walking up to my house", the bare form Lucia already does the job of identifying the individual called Lucia without any further help from articles or suchlike. In fact, "I saw the Lucia walking up to my house" is quite odd (but equivalent sentences are fine in other languages), and proper names in general don't seem to combine easily with modifiers or determiners. 

But then, sometimes they do, and then usually something interesting is happening. In this project we are interested in those cases where proper names do go non-bare and in the semantic and pragmatic effects this has: e.g., are they coerced into a description-like reading, or does this convey some additional attitudinal meaning? Our approach is crosslinguistic: we aim to systematically compare this kind of behaviour across languages, aiming in particular to gather data from typologically diverse and underdescribed languages. The project has so far organized two workshops, one in 2024 in Cologne and another in 2025 in Mexico D.F., with a third taking place in 2026 in Montevideo. We are also working on a questionnaire that will help interested linguists to elicit these kind of data in their language of study. Let us know if you want to participate! Ongoing since 2024. 

Collaboration with Klaus von Heusinger (University of Cologne). 

Demonstratives are used for pointing and attention coordination. We can look at an object in space and say "that" and thereby direct our interlocutor's attention to it. But that also holds in discourse: the "that" in the sentence before the colon refers to parts of the preceding text. In a language like German, demonstratives can also replace personal pronouns, which we most often use when talking about an individual who we've already introduced into the discourse, e.g. a story. So what happens when we use a demonstrative pronoun instead? We pursue the idea that the choice between a personal pronoun (er/sie/es) or one of the two demonstratives in German (der/die/das or dieser/diese/dieses) allows to make reference to discourse structure in a fine-grained way, selecting between candidate referents that have different topical status, e.g. the topic of a whole paragraph, that of just the previous sentence, or one who is in competition with another for the reader's attention. We investigate this using experiments and corpus studies. Some of our results can be found here and here. Ongoing since September 2021.

When we say something, we nearly always convey more content than just what is the "main point" of an utterance. For the main point to be understood, we need a lot of background and secondary information, and some of that can be conveyed alongside, as it were, the main content in the same utterance. Our languages actually provide us with many different devices to mark content as somehow "not main". For example, when I say "Jonas is pissed off that Sören doesn't like to party", I do not assert the proposition that Sören doesn't like to party separately, I rather treat it as if it were established knowledge between my interlocutors and myself. That our languages provide us with ways to do this is extremely useful since it allows us to package content according to how important we think an issue is for the current conversation. But what if we actually want to talk about one of the side issues? An assumption has been that secondary content like this is not accessible - for example for uptake via an anaphoric expression. That seems intuitive at first sight. We test what must happen for secondary content like "that Sören doesn't like to party" to become accessible for anaphora, investigating the effects of factors like discourse continuation and prosody. First results were presented at PaPE 2025 and Xprag 2025. Ongoing since 2024. 

Collaboration with Raúl Bendezu Araujo (University of Potsdam) and Uli Reich (Freie Universität Berlin). 

There are millions of people in Latin America who are fluently bilingual in a local variety of Quechua and a local variety of Spanish. This is a really interesting language pair: for many of the information-structural, discursive and illocutionary meanings for which European varieties of Spanish use intonation, Quechua is described as using morphosyntactic means. In addition, their word-level prosody is quite different. So what happens when speakers who are bilingual in both languages are faced with the same communicative tasks in the two languages? Given that they have resources from both languages at their disposal, will they stick to strategies from just one language at a time? We investigated these questions via a series of dialogical production tasks, collecting about one hour of speech data each from more than 50 speakers of Spanish and Quechua in Huari, Conchucos, Peru, over several months of fieldwork. The resulting annotated speech corpora are some of the first publicly available corpora of their size in any Quechua variety. 

In my PhD thesis (open access here), one of the first quantitative descriptions of Quechua prosody and information structure, I provided evidence that the local variety of Quechua almost does not "care" at all about word stress, but that bilingual speakers produce utterances ranging on a spectrum from regular word accentuation to only phrasal boundary tone marking of larger prosodic groups. I also described how they use a comparatively small intonational repertoire for information structural and illocutionary functions, making more varied use of morphosyntactic means also in their Spanish utterances. 

The published corpora are available at the website of the larger research project at Freie Universität Berlin, which extends similar methodology also to other language pairs spoken in Latin America.

2015-2021.