Conference Agenda

Conceptual change is often described as the revision of earlier understandings in light of new knowledge (Vosniadou, 2013). Yet, a growing body of evidence suggests that intuitive theories are rarely erased; instead, they only become „masked“ and can therefore interfere with later-acquired scientific conceptions (Shtulman & Valcarcel, 2012). Shtulman and Valcarcel (2012) developed a statement-verification paradigm with statements in two types: consistent statements, for which scientific and intuitive theories align (e.g., “Rocks are composed of matter”), and inconsistent statements, for which scientific theory and intuitive conception conflict (e.g., “Air is composed of matter”). Participants typically show lower accuracy and slower responses for theory-inconsistent compared to theory-consistent statements—a phenomenon known as the Conceptual Interference Effect (CIE). Crucially, however, recent work indicates that the CIE is context-dependent: certain contextual embeddings can favor the activation of one representation (intuitive vs. scientific) over another (Shtulman & Young, 2024).

One central dimension of contextual embedding that is highly salient in education—but rarely examined in conceptual interference research—is language. In multilingual learning environments, learners often acquire intuitive conceptions in their first language (L1) through early experience and informal communication, whereas formal science instruction may occur in a second language (L2) (due to, for example, migration contexts, immersion/CLIL, or higher education). Based on the principle of encoding specificity; (Tulving & Thompson, 1973), information is strongly connected to the language of encoding, yielding so-called language-dependent knowledge representations (Marian & Kaushanskaya, 2007. Therefore, retrieval is facilitated when it occurs in the same language context in which the information was encoded (e.g., Grabner et al., 2012). Applied to conceptual change, this implies that intuitive theories may be more strongly bound to L1, whereas scientific theories may be more strongly bound to L2. Consequently, the language of retrieval may affect which representation is more accessible and, in turn, modulate the magnitude of the CIE.

Therefore, the present study tests whether linguistic context modulates interference from intuitive theories. We adapt the statement-verification paradigm introduced by Shtulman and Valcarcel (2012) to a bilingual population in Luxembourg whose everyday experience and early informal learning are primarily Luxembourgish (L1), whereas science instruction is in French (L2). If language acts as a retrieval cue for intuitively acquired conceptions, interference from these intuitive theories should be stronger when reasoning occurs in L1 and reduced when the same judgments are made in L2. We therefore hypothesized a robust CIE (lower accuracy and slower responses for theory-inconsistent than theory-consistent statements) and a Language × Consistency interaction, such that the CIE is smaller in L2 than in L1.

Participants were 37 university students in Luxembourg (25 female; age 18–36) whose L1 was Luxembourgish but who had received formal science instruction in French (L2). The statements developed by Shtulman and Valcarcel were translated into Luxembourgish and French and were presented auditorily. Each participant verified each statement once in Luxembourgish (L1) and once in French (L2). Data were analyzed with two 2×2 repeated-measures ANOVAs with factors Consistency (consistent statements vs. inconsistent statements) and Language (L1 vs. L2).

Accuracy showed a strong main effect of Consistency (F(1,36)=481.70, p<.001): participants were more accurate on theory-consistent than theory-inconsistent statements, replicating the CIE. Critically, the predicted Consistency × Language interaction emerged (F(1,36)=7.05, p=.012): consistency influenced accuracy more strongly in Luxembourgish (L1) than in French (L2), consistent with reduced intuitive-theory interferences in L2. Reaction times mirrored this pattern. There was a significant main effect of Consistency (F(1,36)=111.37, p<.001) and again a significant Consistency × Language interaction (F(1,36)=4.31, p=.045).

Together, these findings support the idea that conceptual interference is not language-invariant: when the language of reasoning matches the language context in which intuitive conceptions were established, those intuitions appear more accessible and thus more competitive during retrieval. Conversely, reasoning in an incongruent language may reduce interference, functioning as a contextual “shield” against intuitively compelling but scientifically incorrect conceptions.

As a pedagogical implication, reasoning in L2 may reduce interference from L1-based intuitions, which could foster learning in scenarios, where interferences are known to be highly detrimental. For CLIL, immersion, and multilingual school systems, these findings motivate more nuanced views of learning and assessment that consider when language congruency is beneficial for intended retrieval (e.g., Saalbach et al., 2013) versus when it may unintentionally strengthen interferences.

While in many regions local dialects have been slowly replaced by standard language, in the north Italian autonomous province of South Tyrol dialect continues to have a high importance in everyday life communication and social interactions (Glaznieks & Frey, 2023). The protection of the distinct languages, German, Italian, Ladin and with all their varieties, has been a focus of the region since its’ annexation to Italy following the first World War, and has been added to the autonomy statute (Das Südtiroler Autonomiestatut, n.d.; Strobl, 2018). However, how dialect affects the exclusion or inclusion of speakers and non-speakers of local linguistic variants remains under-researched. With this ongoing case study, we aim to fill this research gap by investigating how local dialects can become a useful instrument of inclusion or exclusion in society. To do this, we are answering the following research question:

How and to what extent does the knowledge of the local dialect construct an individual´s sense of belonging to the local community?

Our analysis is informed by Bourdieu’s theory of social and cultural capital (1987) and the close connection of cultural capital and a person’s social capital within a community. The study therefore also investigates how cultural capital in the form of linguistic knowledge and self-efficacy affects the participants’ sense of belonging and social cohesion in their respective communities.

To answer this research question, we pursued data from within the community of a German dialect language course through an electronic survey (n = 30 expected), interviews (n = 8 course participants, dialects teachers and administrative staff), and class observations. Composed of both participants from previous courses and the proximate course starting February 2026, as well as the course instructor and administrative staff, this allows us to address the topic of dialect (non-)use and belonging through different lenses within the same community. This enables us to explore how dialect can act as a signifier of belonging and frame for social participation in the South Tyrolian context.

While this study is ongoing, preliminary findings suggest that dialects can be a meaningful tool for informal communication, especially in daily encounters and social situations. Participants further noted that that exclusion can occur non-intentional, through the assumption that the information is unimportant or unnecessary in the given conversation, which can result in a feeling of being unwelcome and inconsiderate. Furthermore, our results show that dialects can also act as a barrier for participation and signifier of outsider-ness for those without adequate self-efficacy in their use. This suggests both the importance of linguistic variety in the South Tyrolian context, but also the need to focus structural resources on dialect language courses for children to adult to gain the linguistic repertoire needed for social participation in this region. The interest of non-dialect speakers in acquiring the local German variation has been observed to be increasing, as noted in early interviews related to the subject.

Our study gives insights into the importance of local dialects for informal social interactions in the region, adding to the academic knowledge about the impact of minority languages on belonging, social cohesion, and social participation. It further offers valuable knowledge on the use of dialects as a tool of inclusion and exclusion in different parts of private and public life. Thus, this study contributes to research on the importance and impact of local dialects for communities in diverse contexts.

Numerical processing in multilingual education is inherently language-dependent (Spelke & Tsivkin, 2001). This has been demonstrated by previous research in the area of arithmetic learning indicating that declarative arithmetic knowledge, meaning our knowledge of arithmetic facts, is language-dependent: switching between the language of encoding and the language of retrieval produces Language Switching Costs (LSC) in the form of poorer performance, both in accuracy and reaction times (Grabner et al., 2012; Hahn et al., 2019, 2025; Volmer et al., 2018). Recent research was able to observe, that these LSC can also occur in the area of procedural arithmetic knowledge, with a mismatch of the language of encoding and the language of retrieval resulting in reaction time drops of around half a second (Baumschlager, in preparation). Despite the already extensive body of research that exists in the area of LSC in arithmetic learning, the underlying cognitive mechanisms behind this phenomenon remain unclear. Models such as the Bilingual Triple Code Model (Lachelin et al., 2023) propose translation processes between the verbal areas of numerical cognition via direct connections among language-dependent numerical systems. In comparison the Encoding-Complex Model (Campbell & Epp, 2004), lacks this direct connection, instead positing a mediating effect of visual code (e.g. Arabic numerals) in transferring numerical information across languages.

To shine a light upon these underlying mechanisms, and to find out about the potential mediating effect of visual code in the transfer of numerical information, we examined multilingual numerical learning and processing in 120 unbalanced bilingual university students. Participants were trained for one hour on a novel arithmetic procedure using either German (L1) number words, English (L2) number words, or Arabic digits. The following day, they were tested on the newly trained procedure in all three codes, yielding a 3 (learning code) × 3 (testing code) design.

A repeated-measures ANOVA revealed a significant interaction between learning code and testing code. Follow-up contrasts indicated directional LSC between learners L1 and L2, with participants exhibiting larger reaction-time differences when being trained in L1 and tested in L2, than when being trained in L2 and tested in L1. Moreover, while switches from the Arabic-digit code to L1 and L2 showed no differences, participants switched more efficiently into the Arabic-digit code when trained in L1 than when trained in L2. This observed interaction was independent both of participants L2 proficiency and bilingual language dominance, as well as uninfluenced by differences in number reading speed of German and English number words.

Taken together, our findings refine our understanding of the relationship between verbal and visual codes in multilingual numerical processing. They suggest an L1-biased asymmetry in the links between multilingual individuals’ verbal and visual codes: access from the visual code to verbal codes is directional, with a stronger connection to learners’ L1. This implies a mediating role of the visual code in translation, producing direction-dependent LSC. The findings of this study are in line with the ideas postulated by the Encoding Complex Model (Campbell & Epp, 2004). Focusing on the educational context, our results highlight a complicating factor in multilingual instruction of mathematics, in the form of the observed L1-biased asymmetry of learners. When teaching mathematics in learners L2, drops in performance due to LSC have to be expected. This effect should be considered in curriculum and assessment design of multilingual learning environments by including extra training time in L2 settings to negate the drops in performance.