Aarhus University Seal

About the project

Danish is an unusual language: because its sound properties tend to obscure the boundaries between words, Danish children appear to be delayed in learning their native tongue, and even adults seem to require more effort to understand one another. This poses fundamental challenges to longstanding theoretical assumptions in the language sciences: that all languages are equally easy to learn and use. In the project "The Puzzle of Danish", we challenge this assumption by studying language processing at the phonological, sentential, and dialogic level of processing in native speakers of Danish.

Preliminary research has revealed some of the difficulties Danish children face during language acquisition, yet we know very little about the impact on adult Danish usage patterns. Through an integrated series of psycholinguistic experiments, we test what Danes hear, understand, and how they communicate. Comparisons are made to Norwegian -- a closely related language with a substantially less "opaque" pronunciation. Investigating the puzzle of Danish has wide-reaching implications not only for the study of language and cognition, but also for society, touching on how we might improve Danish language skills and instruction of L2 learners.

We have been addressing the puzzle of Danish through three interconnected lines of research, investigating the relative brittleness or robustness of the perception of Danish at the phonological (Subproject 1), sentential (Subproject 2), and dialogue (Subproject 3) levels, taking advantage of the natural experimental control afforded by conducting parallel sets of experiments in Danish at Aarhus University (AU) and Norwegian at the University of Bergen (UiB). Below you can read more on the individual subprojects.

Subproject 1

The first subproject of the Puzzle of Danish explores whether the unique sound structure of Danish affects the way Danish speakers process phonological information and how it is different from that of Norwegian speakers. We know little about the impact of the opaque Danish phonology on phonological representations. Given that the ability to form, access, and improve phonological representations plays an important role in the development of both spoken and written language (e.g., Gathercole & Baddeley, 1993), it is important to understand the nature of phonological processing in native Danes. The leading research question for Subproject 1 is therefore: Does the sound structure of Danish affect the nature of phonological representations? And if so: How might this challenge phonological processing and word identification?

To test this, we have carried out series of psycholinguistic experiments built on paradigms such as the Non-word Repetition, Iterated Learning, Rapid Automated Naming (RAN), and Categorical Perception tasks. We collected data from 160 Danish and 160 Norwegian speakers in Denmark and Norway and currently we are collecting 40 more in each of the two countries.

Using these paradigms, we have observed that contextual information plays a more important role for Danes to distinguish between the sounds/words in their native language, than it does for Norwegians' comprehension of their native language. For instance, Danes need longer time than Norwegians to discern between words such as "sendt" and "tændt", when these occur in counterintuitive sentences such as "Hun har tændt en e-mail" eller "Hun har sendt en lampe"! These findings are important for further exploring Danish and Norwegian second language acquisition, as well as clinical contexts. For instance, reliance on context may provide a certain advantage for individuals with hearing loss.

Current research outcomes:

Scientific papers:

  • Ishkhanyan, B., Højen, A., Fusaroli, R., Johansson, C., Tylén, K., & Christiansen, M. H. (2019). Wait for it! Stronger influence of context on categorical perception in Danish than Norwegian. In A. Goel, C. Seifert & C. Freksa (Eds.),  Proceedings of the 41th Annual Conference of the Cognitive Science Society (CogSci 2019). Presented at the 41st Annual Conference of the Cognitive Science Society, Montreal, Quebec. Austin, TX: Cognitive Science Society. https://psyarxiv.com/guzv5/

Presentations:

  • Ishkhanyan, B., Fusaroil, R., Højen, A., Johansson, C., Bleses, D., Tylén, K. & Christiansen, M. H. (2020). The effect of contextual bias on categorical perception: A Danish and Norwegian crosslinguistic study. Talk to be presented at Fonologi i Norden (FiNo) 2020, Dramman, Norway.
  • Ishkhanyan, B., Trecca, F., Højen, A., Bleses, D., Johansson, C., Tylén, K. & Christiansen, M.H. (2020). Cross-linguistic differences in processing unclear stimuli: a Danish and Norwegian study. Poster to be presented at Fonologi i Norden (FiNo) 2020, Dramman, Norway.
  • Ishkhanyan, B., Trecca, F., Højen, A., Bleses, D., Johansson, C., Tylén, K., & Christiansen, M. H. (2019). Has she sent or lit an email? Preliminary results of a Danish and Norwegian categorical perception study. Talk presented at The Seventh Conference of the Scandinavian Association for Language and Cognition, SALC 7.
  • Ishkhanyan, B., Højen, A., Fusaroli, R., Johansson, C., Tylén, K., & Christiansen, M. H. (2019). Wait for it! Stronger influence of context on categorical perception in Danish than Norwegian. Poster presented at the 41st Annual Conference of the Cognitive Science Society, Montreal, Quebec.
  • L.B. Kristensen, F. Trecca, K. Tylén, R. Fusaroli, D. Bleses, A. Højen, C.R. Dideriksen & M.H. Christiansen (2018, February). Effects of Phonetic Reduction on Rapid Automatized Naming. Poster presented at Phonology in the Nordic Countries 2018 (FiNo2018), Lund University, Sweden.
  • F. Trecca, L.B. Kristensen, K. Tylén, R. Fusaroli, D. Bleses, A. Højen, C.R. Dideriksen & M.H. Christiansen (2018, February). Using Non-Word Repetition to Investigate the Nature of Phonological Representations in Adult Speakers of Danish and Norwegian. Poster presented at Phonology in the Nordic Countries 2018 (FiNo2018), Lund University, Sweden.

Subproject 2

We also know very little about whether this contextual dependence at the level of individual sounds and words carries over to the comprehension of whole sentences. We have been testing this using a short story comprehension paradigm in Danish and Norwegian.

In the paradigm, 160 Danes and 160 Norwegians listen to short stories consisting of a preamble (e.g., “the boy walked into the pet store”) and a main event (e.g., “the boy bought a goldfish for the girl”). After each story, they are shown four drawings depicting the characters from the story in different who-did-what-to-whom scenarios (e.g., a boy giving a fish to a girl). They are then asked to click on the picture that matches the story. In some trials, we switch agent and object, creating internal incongruencies in the stories.

We have observed that in such cases of incongruence, Danes are more prone than Norwegians to disregard the actual input and “rectify” the story to a more expectation-driven interpretation: For instance, when hearing the sentence “the goldfish walked into the pet store […] the boy bought a goldfish for the girl”, Danes tend to select the goldfish-gives-boy-to-girl image, in accordance with the preamble! Interestingly, adding acoustic noise to the auditorily presented stories (in order to make the bottom-up signal less informative) seems to induce Norwegians to also changed their processing strategy to rely more on contextual information. Basically, in the presence of noise, Norwegians start behaving more like Danes!

Current research outcomes:

Scientific papers:

  • Trecca, F., Tylén, K., Fusaroli, R., Johansson, C., & Christiansen, M. H. (2019). Top- down information is more important in noisy situations: Exploring the role of pragmatic, semantic, and syntactic information in language processing. In A. Goel, C. Seifert & C. Freksa (Eds.), Proceedings of the 41st Annual Conference of the Cognitive Science Society, 2988-2994. Presented at the 41st Annual Conference of the Cognitive Science Society, Montreal, Quebec. Austin, TX: Cognitive Science Society. https://doi.org/10.31234/osf.io/xp736

Presentations:

  • F. Trecca, K. Tylén, R. Fusaroli, C. Johansson & M.H. Christiansen (2019, July). Top-down information is more important in noisy situations: Exploring the role of pragmatic, semantic, and syntactic information in language processing. Poster to be presented at the 41st Annual Conference of the Cognitive Science Society, Montreal, Quebec.
  • F. Trecca, K. Tylén, R. Fusaroli, C. Johansson & M.H. Christiansen (2019, May). “Gik guldfisken ind i butikken?”: Cross-Scandinavian differences in reliance on top-down information in spoken language processing. Talk to be presented at the 7th Conference of the Scandinavian Association for Language and Cognition, Aarhus University, Denmark.

Subproject 3

The third subproject of the Puzzle of Danish investigates cross-linguistic differences between Danish and Norwegian speakers on a conversational level. Does the opaque sound structure of Danish have an effect on how people interact? Is the mutual understanding between interlocutors under more pressure in Danish, than Norwegian?

To answer these questions we collected a large corpus of 160 conversations in Danish and 160 in Norwegian, systematically varying contextual demands. We developed coding schemes and Natural Language Processing tools to reliably identify and model the occurrence of three conversational mechanisms:

  1. Repair: questions such as “what?” or “who?” that can tell us how often the mutual understanding between people is conflicted.
  2. Backchannels: the use of small words such as “uh-huh” and “okay” that can tell us how often people confirm that their mutual understanding is intact.
  3. Alignment: a measure of how often people repeat words or syntactic structures from previous utterances.

Danes show markedly higher levels of alignment, that is: (1) they repeat their own words, syntax and semantics more than Norwegians; (2) they re-use each other words, syntax and semantics more than Norwegians. This suggests that Danes compensate for the peculiar sound structure of Danish by making their speech more predictable and therefore more robust to misunderstandings. By repeating linguistic structures, Danes continuously update the mutual understanding and makes it easier for each other to be aware of the content of the conversation. These results have the potential be of significant importance to other research areas that is occupied with second language learning and development of natural language processing tools, in particular for Scandinavian languages.

Current research outcomes:

Scientific papers:

  • Dideriksen, C., Fusaroli, R., Tylén, K., Dingemanse, M., & Christiansen, M. H. (2019). Contextualizing Conversational Strategies: Backchannel, Repair and Linguistic Alignment in Spontaneous and Task-Oriented Conversations. In A. Goel, C. Seifert & C. Freksa (Eds.), Proceedings of the 41st Annual Conference of the Cognitive Science Society, 261–267. Presented at the 41st Annual Conference of the Cognitive Science Society, Montreal, Quebec. Austin, TX: Cognitive Science Society.

  • Coding schemes for repair and backchannels: https://bit.ly/2LtUmax
  • NLP tool to automatically and reproducibly analyze lexical, syntactic and semantic alignment: https://pypi.org/project/align/ & https://psyarxiv.com/a5yh9/

  • Open conversational corpora: http://gigaword.dk/

Presentations: 

  • C. Dideriksen, R. Fusaroli, K. Tylén, M. Dingemanse & M.H. Christiansen (2019, July). Contextualizing conversational strategies: Backchannel, repair and linguistic alignment in spontaneous and task-oriented conversations. Paper presented at the 41st Annual Conference of the Cognitive Science Society, Montreal, Quebec.
  • C. Dideriksen, R. Fusaroli, C. Johansson, K. Tylén & M.H. Christiansen (2019, May). Kill or tap? Left or right? Preliminary results from a study on dialogue differences in task oriented and spontaneous conversations. Talk at the 7th Conference of the Scandinavian Association for Language and Cognition, Aarhus University, Denmark.

Why is this important?

These are fundamental psycholinguistic and pragmatic questions concerning perceptual reliance on bottom-up processing of acoustic information vs. top-down inferences based on underspecified cues, contextual sensitivity and prior knowledge. The results seem to underscore the ‘exotic’ nature of Danish as a language that appears to counter the long-held fundamental assumption in the language sciences: that all languages are equally easy to learn and use. By creating publicly available benchmark datasets capturing the unique characteristics of Danish (with Norwegian as control) at the phonological, sentential and dialogue levels of analyses, we provide an empirically broad foundation that others can built on through their own analyses to inform future theoretical debates over the nature language. Our results and datasets can therefore elevate Danish to a special paradigmatic status in the language sciences as an “evidential wedge” with which to challenge current theories of language, paving the way for a new understanding of its fundamental nature. Understanding the Danish puzzle has also important societal implications. Firstly, our results have the potential to provide insights relevant for second-language instruction in Danish. This is of great social significance given that children with a non-western background perform substantially below the mean of monolingual Danish children in academic achievement. Secondly, the project can illuminate Danes’ apparent problems with second-language learning relative to other Scandinavians (Delsing & Åkesson, 2005). Our cross-linguistic comparisons allow us to disentangle effects of phonology from educational practices, which may suggest new ways of improving second-language teaching in Denmark. Thirdly, understanding the Danish phonological representations can provide new evidence to improve general reading instruction in Danish, enhancing Denmark’s worldwide competitiveness. Currently, Denmark has a lower percentage of readers at the top proficiency level compared to the OECD average, and even decreasing from 2000 to 2009 (OECD, 2010).

"The Puzzle of Danish" is supported by a DFF-Research Project Grant from the Danish Council for Independent Research awarded to Professor Morten H. Christiansen.