Word-level prominence and "stress deafness" in Maltese-English bilinguals

Project Members: Maria Lialiou, Anna Bruggeman, Alexandra Vella, Sarah Grech, Petra B. Schumacher, and Martine Grice (CRC 1252 A01, Uni Bielefeld, Uni Malta)

This study investigates “stress deafness” in bilingual speakers of Maltese and Maltese English. Although both reportedly have lexical stress, the acoustic cues
to prominence appear to be relatively weak. Further, word-initial pitch peaks make pitch an unreliable cue to lexical stress, which can be elsewhere in the word.
In a sequence recall task, we show that speakers dominant in Maltese exhibit a classic “stress deafness” effect, similar to speakers of French. Speakers who identify as balanced or Maltese English dominant have more diverse results and do not show such a strong tendency towards “stress deafness”. These speakers may rely on their exposure to other varieties of English to identify (and recall) word prominences. This study suggests that the nature of stress in Maltese might need to be revisited. ICPhS Paper & Poster: https://drive.google.com/drive/folders/1gVOPJcry5B7jWroijH00hTvaqritpWj-?usp=sharing

Evaluative expressions influence prominence

Project members: Umesh Patil, Stefan Hinterwimmer & Petra B. Schumacher (CRC 1252, C05 & C07)

A short discourse in German such as Peter will einen Benz kaufen. Der hat wohl zu viel Geld. (EN: ‘Peter wants to buy a Mercedes-Benz. He apparently has too much money.’), seems to make sense. But if we make a small change to it by replacing the demonstrative pronoun der with another demonstrative pronoun dieser -- Peter will einen Benz kaufen. Dieser hat wohl zu viel Geld. -- it suddenly sounds strange! What makes the first example comprehensible but not the second? We think the first one involves a subjective evaluation of Peter by a narrator which makes it easy for der to refer to Peter, but dieser is unaffected by such evaluation, so it refers to the car Benz and it makes the second discourse sound strange to German speakers (a car obviously cannot have too much money!). We ran two experiments to test if our claim holds water. For the results, check out the presentation "Evaluative expressions influence prominence: effects on die and diese pronouns" by Umesh Patil, Stefan Hinterwimmer and Petra B. Schumacher at the ICPL-III, 2022 (the abstract is here: https://sfb1252.uni-koeln.de/sites/sfb_1252/user_upload/ICPL3/Abstracts_ICPL3/Prominence-Conference-3_Patil_et._al..pdf).

Modeling prominence constraints for German pronouns as weighted retrieval cues

Project members: Umesh Patil & Petra B. Schumacher (CRC 1252, C07)

How are prominence signals such as subjecthood and agenthood realized cognitively? How do German speakers use these signals while processing pronouns such as 'sie' or 'die'? We tried to explain the interaction between these signals through computational modeling. We converted various prominence signals into memory retrieval cues in a cognitive architecture of ACT-R, and tested if the model behaves similar to German speakers. Indeed it does! Check out our work at AMLaP-2022 (ZOOM link to the session: https://virtual.oxfordabstracts.com/#/event/3067/session/50866 and the abstract: https://virtual.oxfordabstracts.com/#/event/3067/submission/199). If you missed that, a shorter presentation (8 mins) from ICCM-2022 is available at: https://mathpsych.org/presentation/800. And if you want to know more details about the work, take a look at our paper at: https://osf.io/cr5pz/.

Investigating German pronouns processing with naturalistic stimuli

Project members: Magdalena Repp, Ingmar Brilmayer & Petra B. Schumacher (CRC 1252, C07)

In this study we investigate the online processing of German pronouns in larger naturalistic discourse context because previous studies have mostly focused on the investigation of highly controlled items. We look at the German d-pronoun der, die in comparison to the personal pronoun er, sie.  First, we conducted a corpus analysis of an excerpt from the German novel Tschick. The corpus study revealed that D-pronouns in the majority of cases refer to subject and agent antecedents and that the antecedents of the d-prons mostly also are d-prons. These finings deviate from prev. literature but can be explained by the dialogue structure of the novel because in direct speech the perspective of the speaker is rendered which allows a d-pon to refer to other referents but the perspectival center.  Subsequently we carried out an ERP experiment where we measured ERPs continuously as participants listened to the audiobook of the novel excerpt. The results revealed a biphasic N400-Late positivity for D-pronouns relative to personal pronouns with respect to expectation-based processing and attentional reorienting. In addition, we found a large frontal positivity for d-pronouns relative to personal pronouns that can be linked to perspective taking (of the main protagonists).

Are rises special in attention orienting? Some evidence from German boundary tones

Project members: Maria Lialiou, Martine Grice, Christine T. Roehr, Ingmar Brilmayer & Petra B. Schumacher (CRC 1252, A01)

This study is about F0 rises in speech and whether they are special in attracting attention. The idea originates from (neuro)cognitive studies reporting that an increase in loudness, pitch, or any other acoustic property in the signal is experienced as looming or approaching by the listener. In our study, we wanted to investigate if this looming also holds in speech, or if it is situated only in the cognitive domain. To achieve this we conducted an EEG experiment in which participants were presented with two auditory oddball lists, one, consisting of a (standard) repetitive sequence of rising intonation occasionally interspersed with a (deviant) falling intonation, and the other one, of the inverse pattern. Rising and falling intonation was realized at the boundary of real (CV.'CV.CV) German words. We are particularly interested in the Mismatched Negative (MMN) and P3 brain responses as they are claimed to index pre-conscious and conscious attentional mechanisms. We hypothesize that if the processing of speech is purely signal based then deviant rising intonation should attract more attention by virtue of it being acoustically more prominent than falling intonation. Yet, given the fact that participants were presented with a list intonation, the repetitive rising sequence might be more natural, as rising intonation marks non-finality in lists, hence listeners might habituate better to this sequence, having thus a stronger reaction to the deviant fall. Our data show that linguistic expectations derived from the context (in our case, list intonation), affect the processing of speech f0 and thus the activation of the attentional mechanisms. Check out our results in AMLaP2022. In the following link, you can access the poster, the abstract, and a quick presentation of the study: https://virtual.oxfordabstracts.com/#/event/3067/submission/217

DOI engenders neural correlates of expectation violations and discourse updating

Project members: Paul Compensis & Petra B. Schumacher

Differential object indexing (DOI) in Bulgarian is a special encoding strategy that is concerned with discourse management and primarily used to mark (perceived) deviances from expectations with respect to the status of referents in discourse. In this ERP experiment, we showed that the presence of DOI directly affects the discourse representations of the referents involved. Referential shifts to less prominent referents were associated with expectation violations and discourse updating, reflected in a N400 - LPS pattern. This study was presented at the 34th CUNY Conference on Human Sentence Processing (you can find a more detailed abstract here https://www.cuny2021.io/2021/02/24/114/). If you want to know more about the project results in general, see https://kups.ub.uni-koeln.de/62893/.

Computational model of German L2 speakers

Project members: Umesh Patil  & Sol Lago

Do second language learners (L2) of German process German differently than German native speakers (L1)? Yes, they do! We can observe the differences in their eye movements during processing German possessive pronouns like "ihre/n". L2 speakers use the gender information slowly and less efficiently than L1 speakers. Interestingly, L2ers make the same mistakes as German L1ers. Can we explain the differences and similarities between L1 and L2 German speakers using a computational model of language processing? In Patil and Lago (2022), we propose two cognitive models of the real-time pronoun processing by L2ers. The models extend an earlier model of L1ers according to two central hypotheses of L2 sentence processing. We compare which model can explain the L1-L2 differences better. If you are interested in more details take a look at our presentation entitled "Computational cognitive modeling of predictive sentence processing in a second language" at this year's Conference on Computational Natural Language Learning (CoNLL-2022) or the preprint of our paper at: https://osf.io/rkuqz/