Learning and Memory

Children's reading and mathematics are two essential abilities that are critical to academic achievements and future career development. Existing neuroimaging research often concentrated on a single ability (reading or mathematical processing) [1][3], or compared them using one specific MRI modality[2][4][5]. However, the multimodal neuroimaging signatures significantly associated with reading and mathematics remained unexplored. Here, via a supervised three-way MRI fusion (fALFF, FA, and GMV), we identified the MRI signatures for comprehensive reading and mathematical processing for 562 children to reveal the common and unique underlying neurobiological basis. Moreover, the longitudinal predictability of the identified baseline MRI signatures for estimating 5 types of cognitive scores one year later (including attention, memory, reasoning, visual perception, and cognitive composite) was examined and validated. 

The variable nature of the brain's response to identical stimuli has long intrigued researchers. The state of arousal appears to be a critical factor given the enhancement of sensory responses during active behaviors, such as whisking or locomotion (1, 2), and through manipulation of the noradrenergic and cholinergic systems (3, 4). However, in absence of overt active behaviors or external arousal modulators, response variability persists, even over time frames of a few seconds (5–7). The nature and source of this seconds-scale fluctuation in the brain's response remains unclear. Recent studies revealed that ongoing spiking activity during rest is organized as a highly structured cascade dynamic that entrains ~70% of neurons across various regions into a temporal sequence of activations spanning multiple seconds (8). Importantly, this cascade dynamic is phase coupled to rapid, seconds-scale arousal modulations, which may in turn contribute to the seconds-scale variance of neural responses. By analyzing large-scale neuronal recordings from mice, we investigated whether the cascade dynamic persists during periods of continuous external sensory stimulation and, if so, whether the dynamic affects how the brain processes sensory inputs. 

Convolutional neural networks (CNNs) have emerged as a popular choice of deep learning architecture for modeling visual processing, as their hierarchical structure and flow of information processing closely resembles the human ventral stream [LeCun, Y. (1989)]. While CNNs have been used to model healthy visual cognition, there remain limitations in biologically plausible in silico modeling of cognitive decline in neurodegenerative diseases, such as Alzheimer's (AD). Previously, we developed methods to simulate neurodegeneration of the visual system through iterative synaptic injury in CNNs [Tuladhar, A. (2021), Moore, J. (2023)]. However, the limitation of CNNs lies in the lack of biologically meaningful learning mechanisms that are similar to cognitive functions, such as memory. These mechanisms are essential for accurately capturing the neuropathogenesis. For instance, the deposition of beta-amyloid peptide and neurofibrillary tangles of tau polymers in the hippocampus leads to cognitive decline in memory tasks among AD patients.

Building on our prior work, in this study, we equipped a CNN with associative memory to enhance biological plausibility, combining two critical cognitive functions of the brain: visual processing of the ventral stream and associative memory of the hippocampus. The model demonstrates intriguing and beneficial properties, including (1) robustness to noisy or occluded image queries and (2) interpretable and sparse representations in network weights. We argue that this model is an improved in silico framework for a healthy brain, as well as the cognitive profiles of AD progression. 

In the era of globalization, people frequently encounter information from other cultures either directly through social interactions or indirectly through social media. In such cases, our cultural mindset is subtly shifted, a process known as cultural learning (Herrmann et al., 2007; Tomasello et al., 1993). Previous research has revealed that individuals can acquire the representations of other cultures, thereby aligning their neurocognitive process more closely with other cultures (Kitayama et al., 2017; Kitayama et al., 2011). However, few studies have investigated this issue in a direct social interaction context, leaving the cognitive process behind naturalistic culture learning and the underlying neural bases unknown. 

In general, we experience surprise when an observation contradicts expectations based on our past experiences. However, we can be surprised for different reasons, depending on the source of our expectations. Expectations can be flexibly drawn from either our general knowledge about how the world works or memories of specific episodes in the past. Because current leading theories on how surprising events are processed do not account for this complexity in sources of predictions, it is unclear if surprise based on different sources of expectations engage the same or distinct neural processes. A currently prominent view is that the hippocampus acts as a comparator between prior experience and incoming information, and plays an important role in detecting events that mismatch our expectations (e.g. Kumaran & Maguire, 2007; Barron, Auksztulewicz, Friston, 2020). However, this idea stems from work in which participants learned arbitrary associations which were violated after minimal delay. For example, violating the temporal order or spatial arrangement of recently experienced items (e.g. Kumaran & Maguire, 2006; Duncan et al., 2012). In short, these studies have demonstrated a role of the hippocampus in processing surprise based on episodic-like memories. However, outside of the laboratory we often rely on our general semantic or schematic knowledge to predict what is likely to happen in a given situation (Elman & McRae, 2019). While some theories predict increased hippocampal engagement when there is a mismatch between prior schematic knowledge and current experience (e.g. SLIMM framework, van Kesteren et al., 2012), there is no direct evidence to support this idea. 

Individual differences in working memory (WM) capacity have been linked to variation in higher order cognition and psychiatric disability. One goal of the NIH Research Domain Criteria (RDoC) paradigm is to formally characterize this relationship and relate neurocognitive markers of WM to psychopathology across broad diagnostic categories [1].
Although a variety of structural and functional brain measures have been shown to account for variance in WM capacity, recent work integrating distinct modalities of brain structure and function explain more variance than can any single modality on its own [2,3]. The present investigation sought to leverage machine learning to characterize the relative importance of different functional and structural neuroimaging measures for predicting WM task performance, trait-level WM capacity, and overall psychiatric disability.