Lifespan Rhythms of Memory and Cognition (RHYME)

Research Scientists
Markus Werkle-Bergner (Project Leader)

Lydia Brundisch (LIFE, since 04/2024)
Elisa Buchberger (until 09/2024)
Ann-Kathrin Joechner (until 12/2024)
Ulman Lindenberger
Tydings McClary (LIFE, since 10/2023)
Chi (Zoe) Ngo (until 06/2025)
Claire Pauley (since 04/2026)
Sarah Power (until 12/2025)
Annika Werwach (MPSCog, since 09/2023)

Early Childhood Memories (ECHO)
Charting the Ontogeny of Memory Component Processes Across Early Childhood (COMIC)
The Interactive Development of Spatial and Episodic Memory Across Childhood (SpaceBrains)
Memory Availability Versus Accessibility in Early Ontogeny Across Species (MEME)
Generalization in Childhood and Adolescence: Statistical Abstraction (G-Casa)

The RHYME project investigates lifespan differences and developmental changes in attention, working memory, and episodic memory at structural, functional, and behavioral levels. It integrates experimental and longitudinal designs, and draws on multimodal neuroimaging data across a broad methodological spectrum. The project closely collaborates with the Hippocampal Circuit and Code for Cognition Lab, which was established in 2021 at the Brain Imaging Centre of the Research Centre for Natural Sciences in Budapest, Hungary, and is led by former Max Planck Partner Group leader Attila Keresztes.

During the reporting period, RHYME's primary efforts were directed at designing and conducting two large-scale, multi-method longitudinal studies, (i) Early Childhood Memories (ECHO), and (ii) Charting the Ontogeny of Memory Component Processes Across Early Childhood (COMIC). Data collection for both studies will be completed in 2026. In addition to these long-term endeavors, the RHYME team also conducted three smaller empirical studies. SpaceBrains examines reciprocal connections between the development of spatial cognition and memory, and is closely aligned to COMIC. Two additional studies, MEME and G-Casa, each supported by early-career grants from the Jacobs Foundation, either extend RHYME's cross-species approach to memory development (MEME), or investigate links between the development of rule-based learning, generalization, and trait anxiety (G-Casa).

Early Childhood Memories (ECHO)

Though infants demonstrate impressive learning capacities, they also show a prevailing inability to recall episodic memories from the first years of life. This phenomenon, known as infantile amnesia, is evolutionarily conserved, but its function, mechanisms, and developmental trajectory remain poorly understood. Traditional theories assume that early memories are either not formed, or, if formed, permanently lost. Recent rodent research, however, indicates that such memories may persist in latent form, rendered inaccessible rather than erased, with subsequent developmental changes in brain anatomy impairing their accessibility. This conundrum of infantile amnesia calls for a systematic, translational approach that integrates rodent and human research.

To shed light on the nature of infantile amnesia, the RHYME team has launched a close collaboration with Tomás Ryan's lab at Trinity College Dublin, which studies infantile amnesia in rodents. Together, they established a coordinated cross-species research effort designed to uncover the processes governing memory persistence and forgetting early in life. Rodent studies enable circuit-level interrogation of memory formation and retrieval, while human infant research characterizes behavioral and EEG signatures of early memory in rich, naturalistic contexts.

To enhance research integration across species, the human studies conducted at the MPI for Human Development use a novel immersive testing environment created for toddlers aged 18 to 24 months, the age period that, according to current assumptions, immediately precedes the onset of infantile amnesia. In a dedicated laboratory space, the team built a 4 × 4 m learning arena with multiple hiding locations and projection-based walls that can rapidly transform into distinct immersive contexts. This setup parallels the contextual demands of rodent maze paradigms (e.g., the Barnes maze) while preserving the ecological validity of real-world memory encoding. Using this setup, toddlers participate in a naturalistic hide-and-seek task in which they learn the location of a hidden toy across varying contexts, supporting the formation of object–location–context associations. All participating children return for initial retrieval 24 hours after the encoding sessions to further consolidate the memory trace. Long-term retention is assessed after one, three, or six months, enabling the construction of a developmental forgetting function close to the alleged infantile amnesia threshold.

To date, the study has enrolled more than 450 toddlers, creating one of the most comprehensive prospective datasets on early memory retention. In a subsample of 110 participants, the team also collected EEG data to identify neural signatures of encoding and retrieval (Dissertation Annika Werwach). Preliminary results reveal substantial individual differences in learning, recall, and forgetting, with indications of memory retention across all ages (see Figure 1). Ongoing analyses apply (i) markerless pose estimation via DeepLabCut to characterize exploration patterns, (ii) trajectory profiling to capture search strategies, and (iii) linguistic and contextual measures to identify developmental predictors of within- and between-person variability in memory formation and retention. Taken together, the initial findings suggest that early memories may be more durable than traditionally assumed, and that changes in accessibility, rather than absolute loss, may characterize infantile amnesia.

*Figure 1. Probing memory performance in late infancy: A human analogue of the Barnes maze used with rodents.* (A) *Task overview.* Schematic illustration of the testing arena, showing two different environments and their corresponding hiding spots (red markers), an example toy character, and the learning procedure. The arena is a closed room whose walls consist of projection screens; scenes can be changed rapidly to create distinct environments. A ceiling-mounted camera system records infants' search behavior and movement trajectories with high temporal resolution. Top-down views of the two environments are shown in (D). (B) *Learning and retrieval procedure.* On Day 1, infants undergo a brief familiarization to experience the arena's features, including the rapid changes between projected environments. They are also introduced to the toy character and the hide-and-seek game. Subsequently, children learn two unique object–context–place associations across multiple repeated trials in two distinct environments (not shown during familiarization). After a 24-hour delay, memory for the components of the learning episodes, as well as for the integrated object–context–location associations, is assessed. Retrieval is repeated after 1, 3, or 6 months to examine long-term retention and forgetting. (C) *Learning performance.* Across both contexts, 18- to 24-month-old children show robust acquisition of the object–context–place associations. In Context A, children reduce random errors (selecting boxes that were never used as hiding spots in either context) and increase correct choices across trials. After switching to Context B, children initially show context errors (choosing the hiding spot associated with Context A) but progressively increase correct selections while maintaining consistently low random-error rates. (E) *Retention and individual differences.* Across age groups and delay conditions, children exhibit generally low levels of forgetting, with some retaining object–context–place associations for up to 6 months. Notably, there are substantial inter-individual differences in initial learning and forgetting. These differences will be further examined using EEG data and analyses of search strategies derived from movement trajectories.

Image: MPI for Human Development

*Figure 1. Probing memory performance in late infancy: A human analogue of the Barnes maze used with rodents.* (A) *Task overview.* Schematic illustration of the testing arena, showing two different environments and their corresponding hiding spots (red markers), an example toy character, and the learning procedure. The arena is a closed room whose walls consist of projection screens; scenes can be changed rapidly to create distinct environments. A ceiling-mounted camera system records infants' search behavior and movement trajectories with high temporal resolution. Top-down views of the two environments are shown in (D). (B) *Learning and retrieval procedure.* On Day 1, infants undergo a brief familiarization to experience the arena's features, including the rapid changes between projected environments. They are also introduced to the toy character and the hide-and-seek game. Subsequently, children learn two unique object–context–place associations across multiple repeated trials in two distinct environments (not shown during familiarization). After a 24-hour delay, memory for the components of the learning episodes, as well as for the integrated object–context–location associations, is assessed. Retrieval is repeated after 1, 3, or 6 months to examine long-term retention and forgetting. (C) *Learning performance.* Across both contexts, 18- to 24-month-old children show robust acquisition of the object–context–place associations. In Context A, children reduce random errors (selecting boxes that were never used as hiding spots in either context) and increase correct choices across trials. After switching to Context B, children initially show context errors (choosing the hiding spot associated with Context A) but progressively increase correct selections while maintaining consistently low random-error rates. (E) *Retention and individual differences.* Across age groups and delay conditions, children exhibit generally low levels of forgetting, with some retaining object–context–place associations for up to 6 months. Notably, there are substantial inter-individual differences in initial learning and forgetting. These differences will be further examined using EEG data and analyses of search strategies derived from movement trajectories.

Image: MPI for Human Development

Charting the Ontogeny of Memory Component Processes Across Early Childhood (COMIC)

Adaptive memory depends on a balance between generalizing across experiences and preserving the unique details of individual events. Computational and neurocognitive models propose that specific experiences are first encoded as hippocampus-dependent episodic memories and gradually become amenable to generalization through consolidation processes, often facilitated by post-learning sleep. Early in development, children show strong abilities to extract schematic knowledge yet struggle to retain detailed episodic information, suggesting a developmental lead–lag relationship between these processes. Understanding this interplay is essential for charting the neural and cognitive underpinnings of memory development.

The COMIC studies investigate the developmental trajectories of generalization and memory specificity across early and middle childhood. In two large-scale data assessments involving nearly 500 children aged 4 to 8 years, RHYME examines how core components of memory emerge, interact, and stabilize (Dissertation Tydings McClary). The longitudinal arm of COMIC tests whether the uneven maturation of intra-hippocampal structures contributes to the late emergence of memory specificity and how sleep-dependent consolidation modulates reliance on rapid generalization. Using an accelerated longitudinal design, 174 children initially aged 4 to 8 years were assessed across three annual measurement points, each comprising four behavioral sessions and structural MRI. First longitudinal analyses across the initial two measurement waves confirm the predicted developmental pattern, with stronger within-person gains in generalization than in memory specificity (see Figure 2).

*Figure 2. Assessing component memory processes in middle childhood.* (A) *Collection Game Task.* Schematic illustration of the task used to assess generalization, context binding, and both conceptual and perceptual specificity in children aged 4–8 years. During encoding, children view a sequence of displays containing a background scene, a central character, and an associated object. Children are told that each character visits different places and collects items for their personal collection. Accordingly, each character is consistently linked to a specific type of background scene (e.g., stages) and a specific class of objects (e.g., musical instruments). Across the encoding phase, each character appears in four intermixed trials, each time paired with a different context and object. During the test phase, four tasks are administered. Generalization is assessed by presenting a character with three objects, only one of which belongs to the relevant object class (e.g., a previously unseen musical instrument). Context binding is measured by re-presenting a character–scene pairing from encoding; children are asked to select the exact object originally shown with that pairing. Conceptual and perceptual specificity are assessed by having children identify the previously encountered object among similar alternatives that vary in their conceptual category or perceptual details. All images in (A) are from flaticon.com. (B) *Longitudinal results.* Analyses across the first two measurement waves reveal steeper developmental gains among younger children in generalization (left panel) than in binding (right panel), the latter reflecting memory specificity. These findings provide the first longitudinal evidence supporting the hypothesis that generalization development precedes detailed memory maturation during early childhood.

Image: MPI for Human Development

*Figure 2. Assessing component memory processes in middle childhood.* (A) *Collection Game Task.* Schematic illustration of the task used to assess generalization, context binding, and both conceptual and perceptual specificity in children aged 4–8 years. During encoding, children view a sequence of displays containing a background scene, a central character, and an associated object. Children are told that each character visits different places and collects items for their personal collection. Accordingly, each character is consistently linked to a specific type of background scene (e.g., stages) and a specific class of objects (e.g., musical instruments). Across the encoding phase, each character appears in four intermixed trials, each time paired with a different context and object. During the test phase, four tasks are administered. Generalization is assessed by presenting a character with three objects, only one of which belongs to the relevant object class (e.g., a previously unseen musical instrument). Context binding is measured by re-presenting a character–scene pairing from encoding; children are asked to select the exact object originally shown with that pairing. Conceptual and perceptual specificity are assessed by having children identify the previously encountered object among similar alternatives that vary in their conceptual category or perceptual details. All images in (A) are from flaticon.com. (B) *Longitudinal results.* Analyses across the first two measurement waves reveal steeper developmental gains among younger children in generalization (left panel) than in binding (right panel), the latter reflecting memory specificity. These findings provide the first longitudinal evidence supporting the hypothesis that generalization development precedes detailed memory maturation during early childhood.

Image: MPI for Human Development

A central theoretical aim of COMIC is to test whether early childhood memory performance reflects a tripartite structure of pattern separation, pattern completion, and generalization. Neurocomputational models posit that these complementary processes enable individuals to retain distinct episodes while building generalizable knowledge. Yet it remains unclear whether these processes are separable early in life or initially manifest as a single, undifferentiated memorability dimension. To address this question, the RHYME team developed a child-appropriate battery of seven tasks yielding 13 behavioral indicators. The experimental design and analytic plan have been accepted as a Stage 1 Registered Report in Developmental Science. Using structural equation modeling, the authors will compare factor structures between younger (4–5 years) and older (6–7 years) children in a sample of 120 participants per group to determine whether and how the three component processes differentiate over development.

Together, the COMIC studies provide a systematic, process-based account of memory development in early to middle childhood. By combining longitudinal, cross-sectional, and neuroimaging methods with developmentally sensitive tasks, the studies will help to delineate how generalization and specificity emerge, interact, and consolidate in maturing brains. The findings will offer a mechanistic foundation for understanding how children are able to form generalizable knowledge that guides behavior in new situations despite difficulties in learning, and eventually retaining, the details of specific episodes.

Key References

Buchberger, E., Brandmaier, A. M., Lindenberger, U., Werkle-Bergner, M., & Ngo, C. T. (in press). The process structure of memory abilities in early and middle childhood [Registered Report]. Developmental Science.

MPG.PuRe

Full

Buchberger, E. S., Joechner, A.-K., Ngo, C. T., Lindenberger, U., & Werkle-Bergner, M. (2024). Age differences in generalization, memory specificity, and their overnight fate in childhood. Child Development, 95(4), e270–e286. https://doi.org/10.1111/cdev.14089

[Elisa S. Buchberger and Ann-Kathrin Joechner contributed equally to this work.].

MPG.PuRe

DOI

publisher-version

supplementary-material

Ngo, C. T., Buchberger, E. S., Nguyen, P. T. U., Newcombe, N. S., & Werkle-Bergner, M. (2024). Building a cumulative science of memory development. Developmental Review, 72, Article 101119. https://doi.org/10.1016/j.dr.2024.101119

MPG.PuRe

DOI

publisher-version

The Interactive Development of Spatial and Episodic Memory Across Childhood (SpaceBrains)

Spatial cognition and episodic memory develop rapidly in childhood and form the basis for higher-order reasoning, academic success, and STEM-related skills. Memory enables children to retrieve past experiences, build structured knowledge, and anticipate future events, while spatial cognition supports exploration, navigation, and the encoding of new information. The two sets of abilities are tightly linked at the neural level, as both rely on the entorhinal–hippocampal system, which supports spatial coding, cognitive mapping, memory specificity, and generalization. Evidence from animal research indicates that entorhinal and hippocampal circuits mature conjointly.

To shed light on corresponding ontogenetic changes in humans, the RHYME team initiated a series of studies examining the maturation of the entorhinal–hippocampal network from early to middle childhood. This work is conducted in close collaboration with Christian Doeller and his team at the MPI for Human Cognitive and Brain Sciences, Leipzig. The collaboration unites complementary expertise and infrastructure, and is supported by the Strategic Innovation Fund of the Max Planck Society. The common goal is to better understand how the refinement of entorhinal grid codes and the maturation of hippocampal subfields jointly scaffold age-related improvements in spatial navigation, memory specificity, and generalization (Dissertation Lydia Brundisch).

To this end, the RHYME team adapted a virtual 3D navigation task for children and adolescents, enabling measurement of accuracy and search strategies across two cross-sectional samples of participants aged 4 to 15 years in a behavioral pilot study (see Figure 3). Currently, fMRI data on a revised version of this task are being collected in children aged 5 to 8 years. Leveraging the longitudinal COMIC cohort, the team integrates grid-code proxies with detailed behavioral and structural MRI data. The expectation is that hexadirectional modulation of BOLD signals will serve as an index of entorhinal grid-like maturity and clarify how hippocampal and cognitive development shape its emergence.

*Figure 3*. *Child-friendly adaptation of a virtual spatial navigation task to assess grid-cell–like signals in human fMRI in 4- to 8-year-old children.* (A) *Task overview*. The first-person view illustrates how participants explore the virtual environment, including distal cues, boundaries, a local cue, and the target location. The top-down view on the right depicts the overall layout of the virtual world and the spatial arrangement of all cues. At the beginning of each trial, participants are shown a cue (a badger or a bear) that indicates which hiding spot they should search for. During the replace phase, children navigate to the memorized hiding spot, after which feedback indicates how close their response was to the correct location (represented by the number of nuts received). During the subsequent collect phase, participants are reminded of the correct hiding spot before moving on to the next trial. (B) *Pilot study results.* Across two pilot studies (N = 75; ages 4–15 years), spatial accuracy (1 – error) increased markedly between ages 4 and 8, whereas performance reached a plateau from approximately age 9 onward. These findings suggest that the period between 4 and 8 years represents a particularly dynamic developmental window for studying the interplay between spatial cognition and memory.

Image: MPI for Human Development

*Figure 3*. *Child-friendly adaptation of a virtual spatial navigation task to assess grid-cell–like signals in human fMRI in 4- to 8-year-old children.* (A) *Task overview*. The first-person view illustrates how participants explore the virtual environment, including distal cues, boundaries, a local cue, and the target location. The top-down view on the right depicts the overall layout of the virtual world and the spatial arrangement of all cues. At the beginning of each trial, participants are shown a cue (a badger or a bear) that indicates which hiding spot they should search for. During the replace phase, children navigate to the memorized hiding spot, after which feedback indicates how close their response was to the correct location (represented by the number of nuts received). During the subsequent collect phase, participants are reminded of the correct hiding spot before moving on to the next trial. (B) *Pilot study results.* Across two pilot studies (N = 75; ages 4–15 years), spatial accuracy (1 – error) increased markedly between ages 4 and 8, whereas performance reached a plateau from approximately age 9 onward. These findings suggest that the period between 4 and 8 years represents a particularly dynamic developmental window for studying the interplay between spatial cognition and memory.

Image: MPI for Human Development

Memory Availability Versus Accessibility in Early Ontogeny Across Species (MEME)

Similar to the ECHO study, the MEME study (PIs: Sarah Power, Zoe Ngo) investigates a central question in cognitive development: Why are episodic memories in early childhood so fragile? Although young children quickly learn about their environment, they struggle to recall recent specific events. It remains unclear whether such early memory failures reflect weak underlying memory traces or difficulties accessing them during retrieval. Rodent work shows that early memories can persist at the engram level despite behavioral forgetting, but an equivalent non-invasive measure in humans is lacking.

To bridge this gap, MEME combines behavioral paradigms with multi-channel EEG in both children and young rodents. In collaboration with Tomás Ryan and his team at Trinity College Dublin, four-year-old children (n = 40) and age-equivalent mice complete parallel object–tone associative learning tasks while undergoing 32-channel EEG during encoding and retrieval. The study focuses on neural activity during the silent post-cue interval to decode the identity of remembered objects, offering a functional index of memory availability. By dissociating availability from accessibility in early childhood, MEME aims to clarify mechanisms of rapid forgetting and lay the foundation for individualized learning approaches tailored to young children's memory profiles.

Generalization in Childhood and Adolescence: Statistical Abstraction (G-Casa)

The G-Casa study (PIs: Zoe Ngo, Ondrej Zika) investigates the development of rule-based learning and generalization in children and adolescents aged 8 to 14 years. It examines how rule dimensionality shapes learning and transfer, and tests whether individual differences in these abilities are associated with trait anxiety. This study is a multi-site collaboration supported by the Jacobs Foundation, with researchers at the MPI for Human Development in Berlin and at University College London.