Research Group Adaptive Memory and Decision Making (2022–2025)
Head: Bernhard Spitzer
Representational Dynamics of Information in Human Working Memory
Adaptive Learning Strategies
The Research Group Adaptive Memory and Decision Making (AMD) was established at the MPI for Human Development in 2022 and relocated to the Technical University (TU) Dresden when Bernhard Spitzer was appointed Chair of Biopsychology in March 2025. During its active period at the MPIB, the research group was funded by an ERC Consolidator Grant, a Heisenberg Award from the Deutsche Forschungsgemeinschaft (DFG), and additional funds from the DFG. The group has also been hosting PhD students from the Max Planck School of Cognition and the International Max Planck Research Schools (IMPRS) LIFE and COMP2PSYCH.
Representational Dynamics of Information in Human Working Memory
The group's main research focus during the reporting period has been on the representational dynamics of information in human working memory (WM). WM processing in the brain has traditionally been considered an exclusive function of the prefrontal cortex but is increasingly found to recruit widespread cortical networks, the precise roles of which remain poorly understood. Moreover, WM storage is anything but static, but highly dynamic and ever-changing. AMD used human neuroimaging, EEG, eye-tracking, and representational geometry analyses to characterize the spatiotemporal dynamics of distributed WM processing in terms of flexible levels of abstraction.
In one series of studies, the group examined the representational dynamics of WM processing during (i) retrospective cueing (attention/inattention), (ii) distraction, and (iii) multi-item maintenance. A first major publication from this line of work used eye-tracking to disclose rapid and adaptive format changes during retro-cued WM storage, where the information's representational format became more generalized and abstract-categorical when attention was temporarily withdrawn from the memorandum (Linde-Domingo & Spitzer, 2024). Another study, using functional imaging, showed that neural WM representation of object orientation was surprisingly generalized (object-independent) even in early visual areas, and coexisted with concrete-visual (object-specific) representational formats within the same regions. A third study, combining EEG and eye-tracking to uncover the temporal dynamics of sustained multi-item maintenance yielded an interesting null finding: unlike predicted by various contemporary theories, the results showed no evidence for rhythmic fluctuations or modulations of WM storage by ongoing EEG oscillations.
A second work package has been concerned with the interplay between momentary WM processing and later recall from long-term memory. Specifically, a set of behavioral experiments tested the hypothesis that WM retrieval may benefit subsequent long-term memory recall, particularly when the WM information was temporarily unattended. The experimental results provided behavioral support for this hypothesis (Born & Spitzer, 2026), providing an important extension to the "testing-effect" phenomenon known in the long-term memory literature. These findings may have implications for applied settings, in that retrieval-based learning strategies appear to be effective even on substantially shorter time scales than previously assumed. In ongoing work, members of the former group continue to investigate the neural mechanisms underlying this phenomenon in WM using functional imaging, EEG, and eye-tracking.
The group's studies on representational WM dynamics are complemented by additional projects on (i) task-related representational reformatting of WM information (eye-tracking and fMRI studies led by IMPRS LIFE student Dilara Zorbek and AMD postdoc Felix Bröhl), (ii) levels of abstraction in artificial neural network models (ANNs; Muttenthaler et al., 2025, and ongoing projects led by AMD PhD students Fabio Bauer and Fabian Kamp), and (iii) neurocomputational differences in WM processing in patients with schizophrenia (work led by AMD PhD student Inés Pont Sanchis, in cooperation with Charité – Universitätsmedizin Berlin). The aforementioned projects are all currently being continued at the Chair of Biopsychology at TUD Dresden University of Technology.
Adaptive Learning Strategies
An additional research focus of the group has been on adaptive learning strategies, particularly during transitive inference learning (e.g., if A>B and B>C, then A>C). Building on the group's seminal modeling work during the previous reporting period (Ciranka et al., 2022), a new study complemented and extended the notion of functional asymmetries in transitive inference learning, where humans appear to predominantly change their beliefs about the winner of a pairwise comparison, while ignoring the loser. Former AMD member Thomas Graham (now PhD student at the Max Planck School of Cognition) showed that this framework accounts not only for behavior in static learning environments but also in settings where latent hierarchies may suddenly change (Graham & Spitzer, 2025). The framework has proven instrumental also in developmental work at the Institute (in collaboration with the Center for Lifespan Psychology). The group's focus on learning and generalization also spurred further cross-group collaboration, e.g., in a project led by the former Max Planck Research Group NeuroCode on the origins of generalization effects in human aversion learning (Verra et al., 2026).
Key References