Tracelink: A model of consolidation and amnesia

A coupled neural field model for the standard consolidation theory

The standard consolidation theory states that short-term memories located in the hippocampus enable the consolidation of long-term memories in the neocortex. In other words, the neocortex slowly learns long-term memories with a transient support of the hippocampus that quickly learns unstable memories. However, it is not clear yet what could be the neurobiological mechanisms underlying these differences in learning rates and memory time-scales. Here, we propose a novel modeling approach of the standard consolidation theory, that focuses on its potential neurobiological mechanisms. In addition to synaptic plasticity and spike frequency adaptation, our model incorporates adult neurogenesis in the dentate gyrus as well as the difference in size between the neocortex and the hippocampus, that we associate with distance-dependent synaptic plasticity. We also take into account the interconnected spatial structure of the involved brain areas, by incorporating the above neurobiological mechanisms in a coupled neural field framework, where each area is represented by a separate neural field with intra- and inter-area connections. To our knowledge, this is the first attempt to apply neural fields to this process. Using numerical simulations and mathematical analysis, we explore the short-term and long-term dynamics of the model upon alternance of phases of hippocampal replay and retrieval cue of an external input. This external input is encodable as a memory pattern in the form of a multiple bump attractor pattern in the individual neural fields. In the model, hippocampal memory patterns become encoded first, before neocortical ones, because of the smaller distances between the bumps of the hippocampal memory patterns. As a result, retrieval of the input pattern in the neocortex at short time-scales necessitates the additional input delivered by the memory pattern of the hippocampus. Neocortical memory patterns progressively consolidate at longer times, up to a point where their retrieval does not need the support of the hippocampus anymore. At longer times, perturbation of the hippocampal neural fields by neurogenesis erases the hippocampus pattern, leading to a final state where the memory pattern is exclusively evoked in the neocortex. Therefore, the dynamics of our model successfully reproduces the main features of the standard consolidation theory. This suggests that neurogenesis in the hippocampus and distance-dependent synaptic plasticity coupled to synaptic depression and spike frequency adaptation, are indeed critical neurobiological processes in memory consolidation.

Alcohol-induced damage to the fimbria/fornix reduces hippocampal-prefrontal cortex connection during early abstinence

INTRODUCTION

Alcohol dependence is characterized by a gradual reduction in cognitive control and inflexibility to contingency changes. The neuroadaptations underlying this aberrant behavior are poorly understood. Using an animal model of alcohol use disorders (AUD) and complementing diffusion-weighted (dw)-MRI with quantitative immunohistochemistry and electrophysiological recordings, we provide causal evidence that chronic intermittent alcohol exposure affects the microstructural integrity of the fimbria/fornix, decreasing myelin basic protein content, and reducing the effective communication from the hippocampus (HC) to the prefrontal cortex (PFC). Using a simple quantitative neural network model, we show how disturbed HC-PFC communication may impede the extinction of maladaptive memories, decreasing flexibility. Finally, combining dw-MRI and psychometric data in AUD patients, we discovered an association between the magnitude of microstructural alteration in the fimbria/fornix and the reduction in cognitive flexibility. Overall, these findings highlight the vulnerability of the fimbria/fornix microstructure in AUD and its potential contribution to alcohol pathophysiology. Fimbria vulnerability to alcohol underlies hippocampal-prefrontal cortex dysfunction and correlates with cognitive impairment.

No evidence for a preferential role of sleep in episodic memory abstraction

Substantial evidence suggests that sleep has a role in declarative memory consolidation. An influential notion holds that such sleep-related memory consolidation is associated with a process of abstraction. The neural underpinnings of this putative process are thought to involve a hippocampo-neocortical dialogue. Specifically, the idea is that, during sleep, the statistical contingencies across episodes are re-coded to a less hippocampus-dependent format, while at the same time losing configural information. Two previous studies from our lab, however, failed to show a preferential role of sleep in either episodic memory decontextualisation or the formation of abstract knowledge across episodic exemplars. Rather these processes occurred over sleep and wake time alike. Here, we present two experiments that replicate and extend these previous studies and exclude some alternative interpretations. The combined data show that sleep has no preferential function in this respect. Rather, hippocampus-dependent memories are generalised to an equal extent across both wake and sleep time. The one point on which sleep outperforms wake is actually the preservation of episodic detail of memories stored prior to sleep.

Randomly fluctuating neural connections may implement a consolidation mechanism that explains classic memory laws

How can we reconcile the massive fluctuations in neural connections with a stable long-term memory? Two-photon microscopy studies have revealed that large portions of neural connections (spines, synapses) are unexpectedly active, changing unpredictably over time. This appears to invalidate the main assumption underlying the majority of memory models in cognitive neuroscience, which rely on stable connections that retain information over time. Here, we show that such random fluctuations may in fact implement a type of memory consolidation mechanism with a stable very long-term memory that offers novel explanations for several classic memory 'laws', namely Jost's Law (1897: superiority of spaced learning) and Ribot's Law (1881: loss of recent memories in retrograde amnesia), for which a common neural basis has been postulated but not established, as well as other general 'laws' of learning and forgetting. We show how these phenomena emerge naturally from massively fluctuating neural connections.

Stochastic consolidation of lifelong memory

Humans have the remarkable ability to continually store new memories, while maintaining old memories for a lifetime. How the brain avoids catastrophic forgetting of memories due to interference between encoded memories is an open problem in computational neuroscience. Here we present a model for continual learning in a recurrent neural network combining Hebbian learning, synaptic decay and a novel memory consolidation mechanism: memories undergo stochastic rehearsals with rates proportional to the memory's basin of attraction, causing self-amplified consolidation. This mechanism gives rise to memory lifetimes that extend much longer than the synaptic decay time, and retrieval probability of memories that gracefully decays with their age. The number of retrievable memories is proportional to a power of the number of neurons. Perturbations to the circuit model cause temporally-graded retrograde and anterograde deficits, mimicking observed memory impairments following neurological trauma.

Modeling Neurotransmission: Computational Tools to Investigate Neurological Disorders

The investigation of synaptic functions remains one of the most fascinating challenges in the field of neuroscience and a large number of experimental methods have been tuned to dissect the mechanisms taking part in the neurotransmission process. Furthermore, the understanding of the insights of neurological disorders originating from alterations in neurotransmission often requires the development of (i) animal models of pathologies, (ii) invasive tools and (iii) targeted pharmacological approaches. In the last decades, additional tools to explore neurological diseases have been provided to the scientific community. A wide range of computational models in fact have been developed to explore the alterations of the mechanisms involved in neurotransmission following the emergence of neurological pathologies. Here, we review some of the advancements in the development of computational methods employed to investigate neuronal circuits with a particular focus on the application to the most diffuse neurological disorders.

Dysregulation of excitatory neural firing replicates physiological and functional changes in aging visual cortex

The mammalian visual system has been the focus of countless experimental and theoretical studies designed to elucidate principles of neural computation and sensory coding. Most theoretical work has focused on networks intended to reflect developing or mature neural circuitry, in both health and disease. Few computational studies have attempted to model changes that occur in neural circuitry as an organism ages non-pathologically. In this work we contribute to closing this gap, studying how physiological changes correlated with advanced age impact the computational performance of a spiking network model of primary visual cortex (V1). Our results demonstrate that deterioration of homeostatic regulation of excitatory firing, coupled with long-term synaptic plasticity, is a sufficient mechanism to reproduce features of observed physiological and functional changes in neural activity data, specifically declines in inhibition and in selectivity to oriented stimuli. This suggests a potential causality between dysregulation of neuron firing and age-induced changes in brain physiology and functional performance. While this does not rule out deeper underlying causes or other mechanisms that could give rise to these changes, our approach opens new avenues for exploring these underlying mechanisms in greater depth and making predictions for future experiments.

A computational model of systems memory consolidation and reconsolidation

In the mammalian brain, newly acquired memories depend on the hippocampus (HPC) for maintenance and recall, but over time, the neocortex takes over these functions, rendering memories HPC-independent. The process responsible for this transformation is called systems memory consolidation. Reactivation of a well-consolidated memory can trigger a temporary return to a HPC-dependent state, a phenomenon known as systems memory reconsolidation. The neural mechanisms underlying systems memory consolidation and reconsolidation are not well understood. Here, we propose a neural model based on well-documented mechanisms of synaptic plasticity and stability and describe a computational implementation that demonstrates the model's ability to account for a range of findings from the systems consolidation and reconsolidation literature. We derive several predictions from the computational model and suggest experiments that may test its validity.

Generalization from episodic memories across time: a route for semantic knowledge acquisition

The storage of input regularities, at all levels of processing complexity, is a fundamental property of the nervous system. At high levels of complexity, this may involve the extraction of associative regularities between higher order entities such as objects, concepts and environments across events that are separated in space and time. We propose that such a mechanism provides an important route towards the formation of higher order semantic knowledge. The present study assessed whether subjects were able to extract complex regularities from multiple associative memories and whether they could generalize this regularity knowledge to new items. We used a memory task in which subjects were required to learn face-location associations, but in which certain facial features were predictive of locations. We assessed generalization, as well as memory for arbitrary stimulus components, over a 4-h post-encoding consolidation period containing wakefulness or sleep. We also assessed the stability of regularity knowledge across a period of several weeks thereafter. We found that subjects were able to detect the regularity structure and use it in a generalization task. Interestingly, the performance on this task increased across the 4hr post-learning period. However, no differential effects of cerebral sleep and wake states during this interval were observed. Furthermore, it was found that regularity extraction hampered the storage of arbitrary facial features, resulting in an impoverished memory trace. Finally, across a period of several weeks, memory for the regularity structure appeared very robust whereas memory for arbitrary associations showed steep forgetting. The current findings improve our understanding of how regularities across memories impact memory (trans)formation.

Structural synaptic plasticity has high memory capacity and can explain graded amnesia, catastrophic forgetting, and the spacing effect

Although already William James and, more explicitly, Donald Hebb's theory of cell assemblies have suggested that activity-dependent rewiring of neuronal networks is the substrate of learning and memory, over the last six decades most theoretical work on memory has focused on plasticity of existing synapses in prewired networks. Research in the last decade has emphasized that structural modification of synaptic connectivity is common in the adult brain and tightly correlated with learning and memory. Here we present a parsimonious computational model for learning by structural plasticity. The basic modeling units are "potential synapses" defined as locations in the network where synapses can potentially grow to connect two neurons. This model generalizes well-known previous models for associative learning based on weight plasticity. Therefore, existing theory can be applied to analyze how many memories and how much information structural plasticity can store in a synapse. Surprisingly, we find that structural plasticity largely outperforms weight plasticity and can achieve a much higher storage capacity per synapse. The effect of structural plasticity on the structure of sparsely connected networks is quite intuitive: Structural plasticity increases the "effectual network connectivity", that is, the network wiring that specifically supports storage and recall of the memories. Further, this model of structural plasticity produces gradients of effectual connectivity in the course of learning, thereby explaining various cognitive phenomena including graded amnesia, catastrophic forgetting, and the spacing effect.

Time, not sleep, unbinds contexts from item memory

Contextual cues are known to benefit memory retrieval, but whether and how sleep affects this context effect remains unresolved. We manipulated contextual congruence during memory retrieval in human volunteers across 12 h and 24 h intervals beginning with either sleep or wakefulness. Our data suggest that whereas contextual cues lose their potency with time, sleep does not modulate this process. Furthermore, our results are consistent with the idea that sleep's beneficial effect on memory retention depends on the amount of waking time that has passed between encoding and sleep onset. The findings are discussed in the framework of competitive consolidation theory.

Synaptic scaling enables dynamically distinct short- and long-term memory formation

Memory storage in the brain relies on mechanisms acting on time scales from minutes, for long-term synaptic potentiation, to days, for memory consolidation. During such processes, neural circuits distinguish synapses relevant for forming a long-term storage, which are consolidated, from synapses of short-term storage, which fade. How time scale integration and synaptic differentiation is simultaneously achieved remains unclear. Here we show that synaptic scaling - a slow process usually associated with the maintenance of activity homeostasis - combined with synaptic plasticity may simultaneously achieve both, thereby providing a natural separation of short- from long-term storage. The interaction between plasticity and scaling provides also an explanation for an established paradox where memory consolidation critically depends on the exact order of learning and recall. These results indicate that scaling may be fundamental for stabilizing memories, providing a dynamic link between early and late memory formation processes.

The role of sleep in forming a memory representation of a two-dimensional space

There is ample evidence from human and animal models that sleep contributes to the consolidation of newly learned information. The precise role of sleep for integrating information into interconnected memory representations is less well understood. Building on prior findings that following sleep (as compared to wakefulness) people are better able to draw inferences across learned associations in a simple hierarchy, we ask how sleep helps consolidate relationships in a more complex representational space. We taught 60 subjects spatial relationships between pairs of buildings, which (unknown to participants) formed a two-dimensional grid. Critically, participants were only taught a subset of the many possible spatial relations, which allowed them to potentially infer the remainder. After a 12 h period that either did or did not include a normal period of sleep, participants returned to the lab. We examined the quality of each participant's map of the two-dimensional space, and their knowledge of relative distances between buildings. After 12 h with sleep, subjects could more accurately map the full space than subjects who experienced only wakefulness. The incorporation of untaught, but inferable, associations was particularly improved. We further found that participants' distance judgment performance related to self-reported navigational style, but only after sleep. These findings demonstrate that consolidation over a night of sleep begins to integrate relations into an interconnected complex representation, in a way that supports spatial relational inference.

A mathematical model of forgetting and amnesia

We describe a mathematical model of learning and memory and apply it to the dynamics of forgetting and amnesia. The model is based on the hypothesis that the neural systems involved in memory at different time scales share two fundamental properties: (1) representations in a store decline in strength (2) while trying to induce new representations in higher-level more permanent stores. This paper addresses several types of experimental and clinical phenomena: (i) the temporal gradient of retrograde amnesia (Ribot’s Law), (ii) forgetting curves with and without anterograde amnesia, and (iii) learning and forgetting curves with impaired cortical plasticity. Results are in the form of closed-form expressions that are applied to studies with mice, rats, and monkeys. In order to analyze human data in a quantitative manner, we also derive a relative measure of retrograde amnesia that removes the effects of non-equal item difficulty for different time periods commonly found with clinical retrograde amnesia tests. Using these analytical tools, we review studies of temporal gradients in the memory of patients with Korsakoff’s Disease, Alzheimer’s Dementia, Huntington’s Disease, and other disorders.

The role of memory consolidation in generalisation of new linguistic information

Accounts of memory that postulate complementary learning systems (CLS) have become increasingly influential in the field of language learning. These accounts predict that generalisation of newly learnt linguistic information to untrained contexts requires offline memory consolidation. Such generalisation should not be observed immediately after training, as these accounts claim unconsolidated representations are context and hippocampus-dependent and gain contextual and hippocampal independence only after consolidation. We trained participants on new affixes (e.g., -nule) attached to familiar word stems (e.g., buildnule), testing them immediately or 2days later. Participants showed an immediate advantage for trained affixes in a speeded shadowing task as long as these affixes occurred in the stem contexts in which they were learnt (e.g., buildnule). This learning effect generalised to words with untrained stems (e.g., sailnule) only in the delayed test condition. By contrast, a non-speeded definition selection task showed immediate generalisation. We propose that generalisation can be supported by initial context-dependent memories given sufficient processing time, but that context-independent lexical representations emerge only following consolidation, as predicted by CLS accounts.

Aging memories: differential decay of episodic memory components

Some memories about events can persist for decades, even a lifetime. However, recent memories incorporate rich sensory information, including knowledge on the spatial and temporal ordering of event features, while old memories typically lack this "filmic" quality. We suggest that this apparent change in the nature of memories may reflect a preferential loss of hippocampus-dependent, configurational information over more cortically based memory components, including memory for individual objects. The current study systematically tests this hypothesis, using a new paradigm that allows the contemporaneous assessment of memory for objects, object pairings, and object-position conjunctions. Retention of each memory component was tested, at multiple intervals, up to 3 mo following encoding. The three memory subtasks adopted the same retrieval paradigm and were matched for initial difficulty. Results show differential decay of the tested episodic memory components, whereby memory for configurational aspects of a scene (objects' co-occurrence and object position) decays faster than memory for featured objects. Interestingly, memory requiring a visually detailed object representation decays at a similar rate as global object recognition, arguing against interpretations based on task difficulty and against the notion that (visual) detail is forgotten preferentially. These findings show that memories undergo qualitative changes as they age. More specifically, event memories become less configurational over time, preferentially losing some of the higher order associations that are dependent on the hippocampus for initial fast encoding. Implications for theories of long-term memory are discussed.

A neural model of hippocampal-striatal interactions in associative learning and transfer generalization in various neurological and psychiatric patients

Building on our previous neurocomputational models of basal ganglia and hippocampal region function (and their modulation by dopamine and acetylcholine, respectively), we show here how an integration of these models can inform our understanding of the interaction between the basal ganglia and hippocampal region in associative learning and transfer generalization across various patient populations. As a common test bed for exploring interactions between these brain regions and neuromodulators, we focus on the acquired equivalence task, an associative learning paradigm in which stimuli that have been associated with the same outcome acquire a functional similarity such that subsequent generalization between these stimuli increases. This task has been used to test cognitive dysfunction in various patient populations with damages to the hippocampal region and basal ganglia, including studies of patients with Parkinson's disease (PD), schizophrenia, basal forebrain amnesia, and hippocampal atrophy. Simulation results show that damage to the hippocampal region-as in patients with hippocampal atrophy (HA), hypoxia, mild Alzheimer's (AD), or schizophrenia-leads to intact associative learning but impaired transfer generalization performance. Moreover, the model demonstrates how PD and anterior communicating artery (ACoA) aneurysm-two very different brain disorders that affect different neural mechanisms-can have similar effects on acquired equivalence performance. In particular, the model shows that simulating a loss of dopamine function in the basal ganglia module (as in PD) leads to slow acquisition learning but intact transfer generalization. Similarly, the model shows that simulating the loss of acetylcholine in the hippocampal region (as in ACoA aneurysm) also results in slower acquisition learning. We argue from this that changes in associative learning of stimulus-action pathways (in the basal ganglia) or changes in the learning of stimulus representations (in the hippocampal region) can have similar functional effects.

Multiple memory stores and operant conditioning: a rationale for memory's complexity

Why does the brain contain more than one memory system? Genetic algorithms can play a role in elucidating this question. Here, model animals were constructed containing a dorsal striatal layer that controlled actions, and a ventral striatal layer that controlled a dopaminergic learning signal. Both layers could gain access to three modeled memory stores, but such access was penalized as energy expenditure. Model animals were then selected on their fitness in simulated operant conditioning tasks. Results suggest that having access to multiple memory stores and their representations is important in learning to regulate dopamine release, as well as in contextual discrimination. For simple operant conditioning, as well as stimulus discrimination, hippocampal compound representations turned out to suffice, a counterintuitive result given findings that hippocampal lesions tend not to affect performance in such tasks. We argue that there is in fact evidence to support a role for compound representations and the hippocampus in even the simplest conditioning tasks.

Topographical, autobiographical and semantic memory in a patient with bilateral mesial temporal and retrosplenial infarction

According to Consolidation Theory (Squire, 1992, Psychological Review, 99, 195; Squire & Alvarez, 1995, Current Opinion in Neurobiology, 5, 169), the mesial temporal lobes have a time-limited role in the maintenance, storage and retrieval of retrograde declarative memories, such that they are not necessary for recalling remote memories. In contrast, proponents of the Multiple Trace Theory (Fuji, Moscovitch, & Nadel, 2000, Handbook of neuropsychology, 2nd ed., p 223, Amsterdam, New York: Elsevier; Nadel & Moscovitch, 1999, Current Opinion in Neurobiology, 7, 217) posit that the mesial temporal lobe (MTL) is necessary for remembering detailed autobiographical and topographical material from all time periods. A third theory of hippocampal function, the Cognitive Map Theory (O'Keefe & Nadel, 1978, The hippocampus as a cognitive map. Oxford: Clarendon), states that the hippocampus is involved in the processing of allocentric spatial representations. The precise role of the MTL in remote memory has been difficult to elucidate, as the majority of studies present cases with widespread brain damage that often occurred many years prior to testing. We investigated retrograde autobiographical, semantic and topographical memories in a subject (SG) who had recently sustained infarctions confined to the MTL and retrosplenial region bilaterally. Inconsistent with the predictions of Cognitive Map Theory, memory for spatial maps that were learned in the past was preserved. Additional testing indicated that SG suffered from a landmark agnosia, which affected remotely and recently acquired information equally. SG was also poor at imagining which direction he would have to turn his body to move from one landmark to another. In accordance with Consolidation Theory, SG performed similarly to control subjects for remote time periods on various measures of retrograde autobiographical memory and demonstrated intact knowledge regarding famous faces and vocabulary terms that were acquired in the past. In contrast, memory for remote public events was impaired. The current findings indicate that the mesial temporal and/or retrosplenial regions have little role to play in memory for remotely acquired spatial maps, autobiographical memories, famous faces or vocabulary terms. However, the findings for landmark naming, directional calculations between landmarks and knowledge of public events suggest that the MTL and/or retrosplenial cortices remain important for accessing these types of memories indefinitely.

Retrograde Amnesia for Autobiographical Memories and Public Events in Mild and Moderate Alzheimer's Disease

Patients with mild to moderate Alzheimer's disease and normal controls were tested on two retrograde memory tests, one based on public events, and the other querying autobiographical memory. On both tests, patients showed strong decrements as compared to normal controls, pointing to retrograde amnesia. Evidence for a gradient in retrograde amnesia was conflicted, with analyses of variance revealing no gradient beyond the most recent period, and more sensitive analyses pointing to shallow Ribot gradients on both tests. A literature review shows that this is the case in most published studies. In autobiographical remote memory patients generated many incorrect answers, a tendency correlated with the number of false alarms on an anterograde memory test administered several months earlier. This suggests a stable, possibly executive, factor underlying memory errors.

Computational models of dementia and neurological problems

A critical goal of neuroscience is to fully understand neural processes and their relations to mental processes, and cognitive, affective, and behavioral disorders. Computational modeling, although still in its infancy, continues to play a central role in this endeavor. Presented here is a review of different aspects of computational modeling that help to explain many features of neuropsychological syndromes and psychiatric disease. Recent advances in computational modeling of epilepsy, cortical reorganization after lesions, Parkinson's and Alzheimer diseases are also reviewed. Additionally, this chapter will also identify some trends in the computational modeling of brain functions.

Methods for reducing interference in the Complementary Learning Systems model: oscillating inhibition and autonomous memory rehearsal

The stability-plasticity problem (i.e. how the brain incorporates new information into its model of the world, while at the same time preserving existing knowledge) has been at the forefront of computational memory research for several decades. In this paper, we critically evaluate how well the Complementary Learning Systems theory of hippocampo-cortical interactions addresses the stability-plasticity problem. We identify two major challenges for the model: Finding a learning algorithm for cortex and hippocampus that enacts selective strengthening of weak memories, and selective punishment of competing memories; and preventing catastrophic forgetting in the case of non-stationary environments (i.e. when items are temporarily removed from the training set). We then discuss potential solutions to these problems: First, we describe a recently developed learning algorithm that leverages neural oscillations to find weak parts of memories (so they can be strengthened) and strong competitors (so they can be punished), and we show how this algorithm outperforms other learning algorithms (CPCA Hebbian learning and Leabra at memorizing overlapping patterns. Second, we describe how autonomous re-activation of memories (separately in cortex and hippocampus) during REM sleep, coupled with the oscillating learning algorithm, can reduce the rate of forgetting of input patterns that are no longer present in the environment. We then present a simple demonstration of how this process can prevent catastrophic interference in an AB-AC learning paradigm.