Working memory (WM) is essential for almost every cognitive task. The neural and synaptic mechanisms supporting the rapid encoding and maintenance of memories in diverse tasks are the subject of an ongoing debate. The traditional view of WM as stationary persistent firing of selective neuronal populations has given room to newer ideas regarding mechanisms that support a more dynamic maintenance of multiple items. Various computational WM models based on different biologically plausible plasticity mechanisms have been proposed. We show that these proposed short-term plasticity mechanisms may not necessarily be competing explanations but instead yield interesting interactions that broaden the functional range of models on a wide set of WM task motifs and simultaneously enhance the biological plausibility of spiking neural network models, in particular of the underlying synaptic plasticity. Although reductionist models (WM function explained by one particular mechanism) are theoretically appealing and have increased our understanding of specific mechanisms, they are narrow explanations. In this study, we evaluate the interactions between three commonly proposed classes of plasticity, namely, intrinsic excitability, synaptic facilitation/augmentation, and Hebbian plasticity. We systematically test combinations of mechanisms in a spiking neural network model on a broad suite of tasks or functional motifs deemed principally important for WM operation, such as one-shot encoding, free and cued recall, and multi-item delay maintenance and updating. Our analysis of the operational task performance indicates that a composite model is superior to more reductionist variants. Importantly, we attribute the observable differences to the principle nature of specific types of plasticity.

The mammalian brain is a complex structure, capable of processing sensory stimuli through the lens of prior knowledge and experiences to guide behavior and decision-making. This process relies on intricate neural dynamics, synaptic plasticity mechanisms, and interactions across brain networks. Despite the brain's remarkable ability to store information, even seemingly stable memories can be modified by new experiences or forgotten over time.

In this work, we use computational modelling to investigate the mechanisms underlying memory functionality, focusing on how the brain supports short- and long-term memory processes. Our research sheds light on episodic, semantic, and working memory phenomena by employing cortical memory models integrating neural plasticity, Bayesian-Hebbian synaptic plasticity across a range of short and long timescales, together with short-term non-Hebbian mechanisms. Inspired by behavioral memory tasks and experimental evidence, we explore processes such as memory semantization — where associated episodic memories are gradually decoupled, allowing for the extraction of abstract semantic meaning. We also investigate and propose hypothetical underlying neurocomputational mechanisms of verbal omissions (memory forgetting) in odor naming tasks. Additionally, we examine the interplay between episodic memory and recency effects in immediate recall. Expanding our framework to working memory, we investigate how different plasticity mechanisms interact to enable both stability and flexibility in memory maintenance.

By bridging computational models with cognitive neuroscience, this research provides new insights into the neural and synaptic basis of memory processes.

Däggdjurshjärnan är en komplex struktur, kapabel att bearbeta sensoriska stimuli genom linsen av tidigare kunskap och erfarenheter och därigenom styra beteende och beslutsfattande. Denna process bygger på intrikat neural dynamik, synaptiska plasticitetsmekanismer och interaktioner mellan hjärnans olika nätverk. Trots hjärnans anmärkningsvärda förmåga att lagra information kan även till synes stabila minnen förändras av nya erfarenheter eller glömmas bort över tid.

I detta arbete använder vi oss av beräkningsmodeller för att undersöka de mekanismer som ligger till grund för olika minnesfunktioner, med fokus på hur hjärnan stödjer minnesprocesser pa lang och kort sikt. Vår forskning belyser fenomen kopplade till episodiskt, semantiskt och arbetsminne genom att använda kortikala minnesmodeller som integrerar neural dynamik, Bayesiansk-Hebbsk synaptisk plasticitet över både korta och långa tidsskalor tillsammans med kortsiktiga icke-Hebbska mekanismer. Inspirerade av olika minnesexperiment och och fynd från dessa undersöker vi processer såsom semantisering av minnet — där associerade episodiska minnen gradvis frikopplas, vilket möjliggör extrahering av abstrakt semantisk innebörd. Vi undersöker också och föreslår hypotetiska underliggande neuroberäkningsmekanismer för verbala utelämnanden (glömning) i luktidentifieringsuppgifter. Dessutom analyserar vi samspelet mellan episodiskt minne och s k recency-effekter i omedelbart minnesåterkallande. Genom att utvidga vår modell till arbetsminnet undersöker vi hur olika plasticitetsmekanismer samverkar för att möjliggöra både stabilitet och flexibilitet i minneslagring.

Genom att förena beräkningsmodeller med kognitiv neurovetenskap bidrar denna forskning med nya insikter om de neurala och synaptiska grunderna för våra minnesprocesser.

We investigated the interaction of episodic memory processes with the short-term dynamics of recency effects. This work takes inspiration from a seminal experimental work involving an odor-in-context association task conducted on rats. In the experimental task, rats were presented with odor pairs in two arenas serving as old or new contexts for specific odor items. Rats were rewarded for selecting the odor that was new to the current context. These new-in-context odor items were deliberately presented with higher recency relative to old-in-context items, so that episodic memory was put in conflict with a short-term recency effect. To study our hypothesis about the major role of synaptic interplay of plasticity phenomena on different time-scales in explaining rats’ performance in such episodic memory tasks, we built a computational spiking neural network model consisting of two reciprocally connected networks that stored contextual and odor information as stable distributed memory patterns. We simulated the experimental task resulting in a dynamic context-item coupling between the two networks by means of Bayesian–Hebbian plasticity with eligibility traces to account for reward-based learning. We first reproduced quantitatively and explained mechanistically the findings of the experimental study, and then to further differentiate the impact of short-term plasticity we simulated an alternative task with old-in-context items presented with higher recency, thus synergistically confounding episodic memory with effects of recency. Our model predicted that higher recency of old-in-context items enhances episodic memory by boosting the activations of old-in-context items. We argue that the model offers a computational framework for studying behavioral implications of the synaptic underpinning of different memory effects in experimental episodic memory paradigms.

Episodic memory is a recollection of past personal experiences associated with particular times and places. This kind of memory is commonly subject to loss of contextual information or “semantization,” which gradually decouples the encoded memory items from their associated contexts while transforming them into semantic or gist-like representations. Novel extensions to the classical Remember/Know (R/K) behavioral paradigm attribute the loss of episodicity to multiple exposures of an item in different contexts. Despite recent advancements explaining semantization at a behavioral level, the underlying neural mechanisms remain poorly understood. In this study, we suggest and evaluate a novel hypothesis proposing that Bayesian–Hebbian synaptic plasticity mechanisms might cause semantization of episodic memory. We implement a cortical spiking neural network model with a Bayesian–Hebbian learning rule called Bayesian Confidence Propagation Neural Network (BCPNN), which captures the semantization phenomenon and offers a mechanistic explanation for it. Encoding items across multiple contexts leads to item-context decoupling akin to semantization. We compare BCPNN plasticity with the more commonly used spike-timing-dependent plasticity (STDP) learning rule in the same episodic memory task. Unlike BCPNN, STDP does not explain the decontextualization process. We further examine how selective plasticity modulation of isolated salient events may enhance preferential retention and resistance to semantization. Our model reproduces important features of episodicity on behavioral timescales under various biological constraints while also offering a novel neural and synaptic explanation for semantization, thereby casting new light on the interplay between episodic and semantic memory processes. 

We present an electrophysiological model of double bouquet cells and integrate them into an established cortical columnar microcircuit model that has previously been used as a spiking attractor model for memory. Learning in that model relies on a Hebbian-Bayesian learning rule to condition recurrent connectivity between pyramidal cells. We here demonstrate that the inclusion of a biophysically plausible double bouquet cell model can solve earlier concerns about learning rules that simultaneously learn excitation and inhibition and might thus violate Dale's principle. We show that learning ability and resulting effective connectivity between functional columns of previous network models is preserved when pyramidal synapses onto double bouquet cells are plastic under the same Hebbian-Bayesian learning rule. The proposed architecture draws on experimental evidence on double bouquet cells and effectively solves the problem of duplexed learning of inhibition and excitation by replacing recurrent inhibition between pyramidal cells in functional columns of different stimulus selectivity with a plastic disynaptic pathway. We thus show that the resulting change to the microcircuit architecture improves the model's biological plausibility without otherwise impacting the model's spiking activity, basic operation, and learning abilities.

We present an electrophysiological model of double bouquet cells and integrate them into an established cortical columnar microcircuit model that has previously been used as a spiking attractor model for memory. Learning in that model relies on a Bayesian-Hebbian learning rule to condition recurrent connectivity between pyramidal cells. We here demonstrate that the inclusion of a biophysically plausible double bouquet cell model can solve earlier concerns about learning rules that simultaneously learn excitation and inhibition and might thus violate Dale's Principle. We show that learning ability and resulting effective connectivity between functional columns of previous network models is preserved when pyramidal synapses onto double-bouquet cells are plastic under the same Hebbian-Bayesian learning rule. The proposed architecture draws on experimental evidence on double bouquet cells and effectively solves the problem of duplexed learning of inhibition and excitation by replacing recurrent inhibition between pyramidal cells in functional columns of different stimulus selectivity with a plastic disynaptic pathway. We thus show that the resulting change to the microcircuit architecture improves the model's biological plausibility without otherwise impacting the models spiking activity, basic operation, and learning abilities.

Odor naming is considered a particularly challenging cognitive test, but the underlying cause of this difficulty is unknown. People often fail to report any source label to identify common odors, resulting in omissions (i.e., a lack of response). Here, with the support of a computational model, we offer a hypothesis about the neural network mechanisms underlying odor naming omissions. Based on an evaluation of behavioral data from almost 40,000 odor naming attempts, we suggest that high omission rates are driven by odors that are referred to by multiple linguistic labels. To explain this observation at the systems level, where olfactory perception and language (semantic) processing are produced by interacting cortical systems, we developed a computational model consisting of two associatively coupled attractor memory networks (odor and language networks), and investigated the effect of Hebbian-like learning on the simulated task performance. We used distributed network representations for the odor percepts and word label mental objects, and accounted for their statistical inter-relationships (correlations) extracted from collected data on odor perceptual similarity, and from a large Swedish odor language corpus, respectively. We evaluated a novel hypothesis, that Bayesian-Hebbian synaptic plasticity mechanisms can explain behavioral omissions in odor naming tasks, casting new light on the underlying mechanisms of this frequently observed memory phenomenon. Due to the nature of Bayesian-Hebbian associative learning connecting the two networks, there was a progressively weaker coupling for odors paired with multiple different labels in the encoding process (one-to-many mapping). Thus, when the model was cued with perceptual odor stimuli that established multiple word label associations (one-to-many mapping), the olfactory language network often produced subthreshold network responses, resulting in elevated omissions (opposite to one-to-few mapping scenario that led to improved performance scores). Our results are of theoretical interest, as they suggest a biologically plausible mechanism to explain a common, but poorly understood, behavioral phenomenon.

We investigated the interaction of episodic memory processes with the short-term dynamics of recency effects. This work takes inspiration from a seminal experimental work involving an odor-in-context association task conducted on rats (Panoz-Brown et al., 2016). In the experimental task, rats were presented with odor pairs in two arenas serving as old or new contexts for specific odor items. Rats were rewarded for selecting the odor that was new to the current context. These new-in-context odor items were deliberately presented with higher recency relative to old-in-context items, so that episodic memory was put in conflict with a short-term recency effect. To study our hypothesis about the major role of synaptic interplay of plasticity phenomena on different time-scales in explaining rats’ performance in such episodic memory tasks, we built a computational spiking neural network model consisting of two reciprocally connected networks that stored contextual and odor information as stable distributed memory patterns. We simulated the experimental task resulting in a dynamic context-item coupling between the two networks by means of Bayesian-Hebbian plasticity with eligibility traces to account for reward-based learning. We first reproduced quantitatively and explained mechanistically the findings of the experimental study, and further simulated an alternative task with old-in-context items presented with higher recency, thus synergistically confounding episodic memory with effects of recency. Our model predicted that higher recency of old-in-context items enhances episodic memory by boosting the activations of old-in-context items. We argue that the model offers a computational framework for studying behavioral implications of the synaptic underpinning of different memory effects in experimental episodic memory paradigms.

Working memory (WM) is essential for almost every cognitive task and behavior. The neural and synaptic mechanisms supporting the rapid encoding and maintenance of memories in diverse tasks are the subject of an ongoing debate. The traditional view of WM as stationary persistent firing of selective neuronal populations has given room to newer ideas regarding mechanisms that support a more dynamic maintenance of multiple items, which may also tolerate more activity disruption. Various computational WM models based on different biologically plausible synaptic and neural plasticity mechanisms have been proposed. We show that these proposed short-term plasticity mechanisms may not necessarily be competing explanations, but instead yield interesting functional interactions on a wide set of WM tasks and enhance the biological plausibility of spiking neural network models, in particular of the underlying synaptic plasticity. While monolithic models (WM function explained by one particular mechanism) are theoretically appealing and have increased our understanding of specific mechanisms, they are narrow explanations. WM models need to become more capable, robust and flexible to account for new experimental evidence of bursty and activity-silent multi-item maintenance in more challenging WM tasks, and generally solve more than one particular task. More detailed models also allow for electrophysiological constraints from recordings.

In this study we evaluate the interactions between three commonly used classes of plasticity, namely intrinsic excitability, synaptic facilitation/augmentation and Hebbian plasticity. Combinations of these are systematically tested in a spiking neural network model on a broad suite of tasks or functional motifs deemed principally important for WM operation, such as one-shot encoding, free and cued recall, delay maintenance and updating. In our evaluation we focus on the operational task performance and biological plausibility. Our results indicate that a composite model, combining several commonly proposed plasticity mechanisms for WM function, is superior to more reductionist variants. Importantly, we attribute the observable differences to the principle nature of specific types of plasticity.