Research

The project is a subproject of the DFG research group "Lasting Learning: Cognitive Mechanisms and Effective Instructional Implementation". It consists of two parts, one of which is situated at the University of Kassel (PI Ralf Rummer), while the other is situated at the University of Passau (PI Judith Schweppe).

The (backward) testing effect constitutes the finding that practicing retrieval after initial study is more effective than other restudy techniques. The effect is one of the best-researched findings on long-term learning and has been demonstrated in both lab and classroom studies. However, retrieval practice with more complex materials might not be beneficial for all learning. Indeed, several studies using complex materials have failed to find a testing effect. If this assumption is true, its applicability in school and higher education would clearly be compromised. Conversely, if retrieval practice is prematurely discarded as beneficial for more complex materials, learners would be deprived of one of the most effective techniques for lasting learning, especially seniors in high school and students in higher education who are typically learning more complex subject matter. The main goal of this research project is to test the alternative explanation that learners had not adequately understood the subject matter by the time the consolidation phase started, which resulted in less effective retrieval practice than with less complex materials. Thus, we assume that the testing effect will occur even with complex learning materials when learners adequately understand the learning material in the initial learning phase. We plan two series of field experiments and an additional joint experiment that will take place in eleventh-grade German classes, and the learning materials will cover a topic that is part of the curriculum.

This project aims to support learners individually when learning both course contents and additional skills by employing digital AI-supported tools (see https://komp-hi.de/). Our subproject focuses on applying common ground processes to interactions between humans and digital learning tools.

Speakers typically adapt their utterances to the (estimated) knowledge of listeners. Specifically, speakers usually avoid giving information that is already known to listeners and at the same time take care to give necessary information that listeners might not know. Communication is most effective when communication partners have a large shared knowledge base, called common ground. In higher education, grounding, that is building a common ground, is an important part of teaching. However, at the beginning, common ground is typically low. Moreover, large courses prevent adapting the teaching individually to students. AI-supported digital learning tools can help grounding by adapting to individual prior knowledge. The subproject Common Ground uses digital tools, currently focusing on chatbots, to support grounding. We develop chatbots that use common ground principles known from communication psychology to aid students learn course materials. Additionally, we test whether these chatbots facilitate learning compared to other learning modes.

This project aims at the investigation of potential moderators (mood, disfluency) of the "seductive details" effect, which indicates that the presentation of information that attracts attention but does not promote comprehension of the learning material minimizes learning efficiency. Dependent variables will be both the learning efficiency and eye movements. We expect higher learning efficiency via a reduction of distractibility when participants are in a bad mood and the text is more difficult to read. The respective eye-tracking experiments are based on the modified paradigm of Ketzer-Nöltge et al. (2019).

The overarching objective of the present project is to gain a deeper understanding of valence sound symbolism, the phenomenon that lip-spreading (vs. lip-rounding) vowels are associated with positive (vs. negative) affective valence. For example, when asked to invent pseudo-names for faces, participants tend to more frequently use /i/ for happy faces and /o/ for angry faces (e.g., Körner & Rummer, 2023). The present project is intended to answer two questions concerning valence sound symbolism.

First, is valence sound symbolism so deeply engrained as to lead to more efficient processing? We will examine automaticity of valence sound symbolism by measuring response speed and memory performance for matching (e.g., /i/ and a happy face) compared with mismatching (e.g., /i/ and an angry face) stimuli. If matching stimuli facilitate performance, this will support the hypothesis that valence sound symbolism is deeply engrained in the human mind and may influence online language processing.

Second, is valence sound symbolism—in contrast to brightness and size sound symbolism—driven by an articulatory (vs. an acoustic) mechanism? Size and brightness are both associated with pitch and might therefore rest on vowel frequency. For valence sound symbolism, in contrast, we have found initial evidence that facial muscle tension employed in vowel articulation is the driving cognitive mechanism (e.g., Körner & Rummer, 2022). Accordingly, we hypothesize that their underlying cognitive mechanisms differ.

The work program of the current research project was twofold. On a methodical level, it aimed at establishing a comprehensive, state-of-the-art EEG laboratory in the University of Kassel’s Department of Psychology. To this end, a powerful research EEG system was installed, together with several sound-attenuating testing booths, in a newly renovated laboratory space. The project encompassed the creation, critical examination, and optimization of data acquisition protocols as well as the practical training of qualified student assistants and the securing of necessary ethical approvals. These measures were then put to the test in the context of several EEG studies that have so far been conducted successfully in the established research laboratory. In terms of content, the project was concerned with the EEG-based investigation of sound-symbolic associations in the human brain, thereby targeting the neuronal basis of a historically disputed principle of the organization of human language. In an ongoing series of experiments, we analyze(d) the neuronal processing of sound-symbolically matching (vs. mismatching) object–sound pairs, concerning for example the pairing between different product categories and different fictional brand names (Glim & Hillje, 2024). In accordance with our hypotheses, information that followed sound-symbolic principles has (so far) been shown to be associated with a facilitation of neuronal processing—a finding suggesting that the concept of sound symbolism is deeply anchored in the human brain, rather than being merely a curious byproduct of the evolution of language.

Sound symbolism refers to the phenomenon that certain aspects of word form covary with certain dimensions of word meaning. Within the scope of this project, three sound-symbolic effects were investigated for which the vowels /i/ and /o/ are of relevance: (1) the relation between /i/ versus /o/ and the emotional valence of the to-be-named objects; (2) the relation between /i/ versus /o/ and the size of the to-be-named objects; and (3) the relation between /i/ versus /o/ and the form attributes round versus angular. In order to test the respective relation-centered hypotheses, we presented positively versus negatively valenced objects (experiment 1), large or small objects (experiment 2) as well as round versus angular forms (experiment 3). The participants' task was to name these objects in a fantasy language. The dependent variable was the number of /i/s and /o/s in the fantasy names. We assumed that the fantasy names for positive (vs. negative) objects contained more /i/s and less /o/s. The same applied to small (vs. large) and angular (vs. round) objects.

The grammatically masculine form in German is intrinsically ambiguous, referring either to male people specifically or to people of any gender generically. Behavioral studies have suggested that this dual function leads to a so-called male bias in elicited mental representations. The underlying neuro-cognitive mechanisms related to such a bias, however, have so far been unexplored. In a total of three EEG-ERP studies (Glim et al., 2023, 2024, 2025), we analyzed the neuronal processing of anaphoric references to men versus women following a role noun in one of several different gender forms in German. The collected data showed that the generic masculine (“Studenten”) leads to recurring processing difficulties when female (rather than male) referents are encountered, both during early perceptual processing and during higher-order reference resolution. In addition, we demonstrated that neither the feminine–masculine pair form (“Studentinnen und Studenten”) nor the gender star form (“Student*innen”) elicits gender-balanced mental representations, with both forms obstructing the neuronal processing of references to men (vs. women), albeit in varying spatio-temporal processing ranges. The current research project was the first to provide comprehensive neurophysiological data on the processing of different gender forms in German and, by doing so, helped put the ongoing societal debate on gender-fair language on robust empirical grounds.

Word form has been found to be systematically associated with word meaning, a phenomenon termed sound symbolism. For example, it has been shown that /i/ (as in meet) is associated with positive valence whereas rounded vowels (such as /o/ and /u/) are associated with negative valence (Rummer et al., 2014). As in the preceding example, research typically concentrates on individual letters (graphemes or phonemes) as target word form features. The present project, in contrast, examined a larger-grained word form feature, specifically word form difficulty. We found that form difficulty is associated with auditory (vs. other sensory) concepts, so that difficult (vs. easy) word forms were more likely judged suitable for auditory concepts.