Unstable Memories Create a High-Level Representation that Enables Learning Transfer

Neuroscience: Memory modification without catastrophe

Adaptive behaviour is supported by changes in neuronal networks. Insight into maintaining these memories - preventing their catastrophic loss - despite further network changes occurring due to novel learning is provided in a new study.

Generalization of procedural motor sequence learning after a single practice trial

When humans begin learning new motor skills, they typically display early rapid performance improvements. It is not well understood how knowledge acquired during this early skill learning period generalizes to new, related skills. Here, we addressed this question by investigating factors influencing generalization of early learning from a skill A to a different, but related skill B. Early skill generalization was tested over four experiments (N = 2095). Subjects successively learned two related motor sequence skills (skills A and B) over different practice schedules. Skill A and B sequences shared ordinal (i.e., matching keypress locations), transitional (i.e., ordered keypress pairs), parsing rule (i.e., distinct sequence events like repeated keypresses that can be used as a breakpoint for segmenting the sequence into smaller units) structures, or possessed no structure similarities. Results showed generalization for shared parsing rule structure between skills A and B after only a single 10-second practice trial of skill A. Manipulating the initial practice exposure to skill A (1 to 12 trials) and inter-practice rest interval (0-30 s) between skills A and B had no impact on parsing rule structure generalization. Furthermore, this generalization was not explained by stronger sensorimotor mapping between individual keypress actions and their symbolic representations. In contrast, learning from skill A did not generalize to skill B during early learning when the sequences shared only ordinal or transitional structure features. These results document sequence structure that can be very rapidly generalized during initial learning to facilitate generalization of skill.

Distinct frequencies balance segregation with interaction between different memory types within a prefrontal circuit

Once formed, the fate of memory is uncertain. Subsequent offline interactions between even different memory types (actions versus words) modify retention.^1,2,3,4,5,6 These interactions may occur due to different oscillations functionally linking together different memory types within a circuit.^{7,8,9,10,11,12,13} With memory processing driving the circuit, it may become less susceptible to external influences.¹⁴ We tested this prediction by perturbing the human brain with single pulses of transcranial magnetic stimulation (TMS) and simultaneously measuring the brain activity changes with electroencephalography (EEG^15,16,17). Stimulation was applied over brain areas that contribute to memory processing (dorsolateral prefrontal cortex, DLPFC; primary motor cortex, M1) at baseline and offline, after memory formation, when memory interactions are known to occur.^1,4,6,10,18 The EEG response decreased offline (compared with baseline) within the alpha/beta frequency bands when stimulation was applied to the DLPFC, but not to M1. This decrease exclusively followed memory tasks that interact, revealing that it was due specifically to the interaction, not task performance. It remained even when the order of the memory tasks was changed and so was present, regardless of how the memory interaction was produced. Finally, the decrease within alpha power (but not beta) was correlated with impairment in motor memory, whereas the decrease in beta power (but not alpha) was correlated with impairment in word-list memory. Thus, different memory types are linked to different frequency bands within a DLPFC circuit, and the power of these bands shapes the balance between interaction and segregation between these memories.

Cross-Modal Facilitation of Episodic Memory by Sequential Action Execution

Throughout our lives, the actions we produce are often highly familiar and repetitive (e.g., commuting to work). However, layered upon these routine actions are novel, episodic experiences. Substantial research has shown that prior knowledge can facilitate learning of conceptually related new information. But despite the central role our behavior plays in real-world experience, it remains unclear how engagement in a familiar sequence of actions influences memory for unrelated, nonmotor information coincident with those actions. To investigate this, we had healthy young adults encode novel items while simultaneously following a sequence of actions (key presses) that was either predictable and well-learned or random. Across three experiments (N = 80 each), we found that temporal order memory, but not item memory, was significantly enhanced for novel items encoded while participants executed predictable compared with random action sequences. These results suggest that engaging in familiar behaviors during novel learning scaffolds within-event temporal memory, an essential feature of episodic experiences.

The Effect of Different Combinations of Practice Schedules on Motor Response Stability during Practice

Many results in motor learning have indicated that relative and absolute timing dimensions are modulated by factors that modify response stability among trials. One of these factors is the combination of constant and variable practices. Although many researchers have investigated the combination of practice schedules, these researchers have used measurements that do not assess performance and motor response separately. This study aimed to investigate the effect of different combinations of practice schedules on motor response stability during practice. Participants performed a sequential key-pressing task with two goals: (1) to learn the relative timing dimension and (2) the absolute timing dimension. Participants were assigned to one of two groups: constant-variable or variable-constant. Our findings indicate an influence of the increase in variability over the practice in the constant-variable group. Precisely, the increase in variability of total time in the second half (constant-variable group) of practice was followed by the maintenance of the same level of cross-correlate between absolute timing error and variability of total time. Finally, our findings support the hypothesis that practicing in a constant schedule favors the relative timing dimension of learning regardless of the order in which the constant practice is provided.

Forgetting Enhances Episodic Control With Structured Memories

Forgetting is a normal process in healthy brains, and evidence suggests that the mammalian brain forgets more than is required based on limitations of mnemonic capacity. Episodic memories, in particular, are liable to be forgotten over time. Researchers have hypothesized that it may be beneficial for decision making to forget episodic memories over time. Reinforcement learning offers a normative framework in which to test such hypotheses. Here, we show that a reinforcement learning agent that uses an episodic memory cache to find rewards in maze environments can forget a large percentage of older memories without any performance impairments, if they utilize mnemonic representations that contain structural information about space. Moreover, we show that some forgetting can actually provide a benefit in performance compared to agents with unbounded memories. Our analyses of the agents show that forgetting reduces the influence of outdated information and states which are not frequently visited on the policies produced by the episodic control system. These results support the hypothesis that some degree of forgetting can be beneficial for decision making, which can help to explain why the brain forgets more than is required by capacity limitations.

Neural inhibition for continual learning and memory

•

In humans, the concentration of cortical GABA decreases during learning.

•

Inhibitory plasticity provides homeostatic control to restore network stability.

•

Memories are held dormant unless latent inhibitory connections are unmasked.

•

Cortical inhibition protects overlapping memories from interference.

•

The emerging model suggests neural inhibition promotes continual learning.

Humans are able to continually learn new information and acquire skills that meet the demands of an ever-changing environment. Yet, this new learning does not necessarily occur at the expense of old memories. The specialised biological mechanisms that permit continual learning in humans and other mammals are not fully understood. Here I explore the possibility that neural inhibition plays an important role. I present recent findings from studies in humans that suggest inhibition regulates the stability of neural networks to gate cortical plasticity and memory retrieval. These studies use non-invasive methods to obtain an indirect measure of neural inhibition and corroborate comparable findings in animals. Together these studies reveal a model whereby neural inhibition protects memories from interference to permit continual learning. Neural inhibition may, therefore, play a critical role in the computations that underlie higher-order cognition and adaptive behaviour.

Bidirectional competitive interactions between motor memory and declarative memory during interleaved learning

Distinct motor and declarative memory systems are widely thought to compete during memory consolidation and retrieval, yet the nature of their interactions during learning is less clear. Recent studies have suggested motor learning not only depend on implicit motor memory system supporting gradual tuning of responses by feedback but also depend on explicit declarative memory system. However, this competition has been identified when both systems are engaged in learning the same material (motor information), and so competition might be emphasized. We tested whether such competition also occurs when learning involved separate motor memory and declarative information presented distinctly but yet in close temporal proximity. We measured behavioral and brain-activity correlates of motor-declarative competition during learning using a novel task with interleaved motor-adaptation and declarative-learning demands. Despite unrelated motor versus declarative information and temporal segregation, motor learning interfered with declarative learning and declarative learning interfered with motor learning. This reciprocal competition was tightly coupled to corresponding reductions of fMRI activity in motor versus declarative learning systems. These findings suggest that distinct motor and declarative learning systems compete even when they are engaged by system-specific demands in close temporal proximity during memory formation.

A Common Task Structure Links Together the Fate of Different Types of Memories

Our memories frequently have features in common. For example, a learned sequence of words or actions can follow a common rule, which determines their serial order, despite being composed of very different events [1, 2]. This common abstract structure might link the fates of memories together. We tested this idea by creating different types of memory task: a sequence of words or actions that either did or did not have a common structure. Participants learned one of these memory tasks and then they learned another type of memory task 6 h later, either with or without the same structure. We then tested the newly formed memory's susceptibility to interference. We found that the newly formed memory was protected from interference when it shared a common structure with the earlier memory. Specifically, learning a sequence of words protected a subsequent sequence of actions learned hours later from interference, and conversely, learning a sequence of actions protected a subsequent sequence of words learned hours later from interference provided the sequences shared a common structure. Yet this protection of the newly formed memory came at a cost. The earlier memory had disrupted recall when it had the same rather than a different structure to the newly formed and protected memory. Thus, a common structure can determine what is retained (i.e., protected) and what is modified (i.e., disrupted). Our work reveals that a shared common structure links the fate of otherwise different types of memories together and identifies a novel mechanism for memory modification.

Memories replayed: reactivating past successes and new dilemmas

Our experiences continue to be processed 'offline' in the ensuing hours of both wakefulness and sleep. During these different brain states, the memory formed during our experience is replayed or reactivated. Here, we discuss the unique challenges in studying offline reactivation, the growth in both the experimental and analytical techniques available across different animals from rodents to humans to capture these offline events, the important challenges this innovation has brought, our still modest understanding of how reactivation drives diverse synaptic changes across circuits, and how these changes differ (if at all), and perhaps complement, those at memory formation. Together, these discussions highlight critical emerging issues vital for identifying how reactivation affects circuits, and, in turn, behaviour, and provides a broader context for the contributions in this special issue. This article is part of the Theo Murphy meeting issue 'Memory reactivation: replaying events past, present and future'.

Visual-oculomotor interactions facilitate consolidation of perceptual learning

Visual skill learning is commonly considered a manifestation of brain plasticity. Following encoding, consolidation of the skill may result in between-session performance gains. A great volume of studies have demonstrated that during the offline consolidation interval, the skill is susceptible to external inputs that modify the preformed representation of the memory, affecting future performance. However, while basic visual perceptual learning is thought to be mediated by sensory brain regions or their higher-order readout pathways, the possibility of visual-oculomotor interactions affecting the consolidation interval and reshaping visual learning remains uncharted. Motivated by findings mapping connections between oculomotor behavior and visual performance, we examined whether visual consolidation can be facilitated by visual-oculomotor interactions. To this aim, we paired reactivation of an oculomotor memory with consolidation of a typical visual texture discrimination task. Importantly, the oculomotor memory was encoded by learning of the pure motor component of the movement, removing visual cues. When brief reactivation of the oculomotor memory preceded the visual task, visual gains were substantially enhanced compared with those achieved by visual practice per se and were strongly related to the magnitude of oculomotor gains, suggesting that the brain utilizes oculomotor memory to enhance basic visual perception.

The role of executive control in post-stroke aphasia treatment

Executive control (EC) ability is increasingly emerging as an important predictor of post-stroke aphasia recovery. This study examined whether EC predicted immediate treatment gains, treatment maintenance and generalization after naming therapy in ten adults with mild to severe chronic post-stroke aphasia. Performance on multiple EC tasks allowed for the creation of composite scores for common EC, and the EC processes of shifting, inhibition and working memory (WM) updating. Participants were treated three times a week for five weeks with a phonological naming therapy; difference scores in naming accuracy of treated and untreated words (assessed pre, post, four- and eight-weeks after therapy) served as the primary outcome measures. Results from simple and multiple linear regressions indicate that individuals with better shifting and WM updating abilities demonstrated better maintenance of treated words at four-week follow-up, and those with better common EC demonstrated better maintenance of treated words at both four- and eight-week follow-ups. Better shifting ability also predicted better generalization to untreated words post-therapy. Measures of EC were not indicative of improvements on treated words immediately post-treatment, nor of generalization to untreated words at follow-up. Findings suggest that immediate treatment gains, maintenance and generalization may be supported by different underlying mechanisms.

The Hippocampus and Neocortical Inhibitory Engrams Protect against Memory Interference

Our experiences often overlap with each other, yet we are able to selectively recall individual memories to guide decisions and future actions. The neural mechanisms that support such precise memory recall remain unclear. Here, using ultra-high field 7T MRI we reveal two distinct mechanisms that protect memories from interference. The first mechanism involves the hippocampus, where the blood-oxygen-level-dependent (BOLD) signal predicts behavioral measures of memory interference, and representations of context-dependent memories are pattern separated according to their relational overlap. The second mechanism involves neocortical inhibition. When we reduce the concentration of neocortical GABA using trans-cranial direct current stimulation (tDCS), neocortical memory interference increases in proportion to the reduction in GABA, which in turn predicts behavioral performance. These findings suggest that memory interference is mediated by both the hippocampus and neocortex, where the hippocampus separates overlapping but context-dependent memories using relational information, and neocortical inhibition prevents unwanted co-activation between overlapping memories.

•

Memory interference is mediated by both the hippocampus and neocortical inhibition

•

In the hippocampus, overlapping memories are separated using relational information

•

In neocortex, inhibition protects overlapping memories from interference

•

By reducing neocortical GABA with brain stimulation, memory interference increases

Memory instability as a gateway to generalization

Our present frequently resembles our past. Patterns of actions and events repeat throughout our lives like a motif. Identifying and exploiting these patterns are fundamental to many behaviours, from creating grammar to the application of skill across diverse situations. Such generalization may be dependent upon memory instability. Following their formation, memories are unstable and able to interact with one another, allowing, at least in principle, common features to be extracted. Exploiting these common features creates generalized knowledge that can be applied across varied circumstances. Memory instability explains many of the biological and behavioural conditions necessary for generalization and offers predictions for how generalization is produced.

Dual enhancement mechanisms for overnight motor memory consolidation

Our brains are constantly processing past events [1]. These off-line processes consolidate memories, leading in the case of motor skill memories to an enhancement in performance between training sessions. A similar magnitude of enhancement develops over a night of sleep following an implicit task, when a sequence of movements is acquired unintentionally, or following an explicit task, when the same sequence is acquired intentionally [2]. What remains poorly understood, however, is whether these similar offline improvements are supported by similar circuits, or through distinct circuits. We set out to distinguish between these possibilities by applying Transcranial Magnetic Stimulation (TMS), over the primary motor cortex (M1) or the inferior parietal lobule (IPL) immediately after learning in either the explicit or implicit task. These brain areas have both been implicated in encoding aspects of a motor sequence, and subsequently supporting offline improvements over sleep [3-5]. Here we show that offline improvements following the explicit task are dependent upon a circuit that includes M1 but not IPL. By contrast, offline improvements following the implicit task are dependent upon a circuit that includes IPL but not M1. Our work establishes the critical contribution made by M1 and IPL circuits to offline memory processing, and reveals that distinct circuits support similar offline improvements.

Statistical learning leads to persistent memory: Evidence for one-year consolidation

Statistical learning is a robust mechanism of the brain that enables the extraction of environmental patterns, which is crucial in perceptual and cognitive domains. However, the dynamical change of processes underlying long-term statistical memory formation has not been tested in an appropriately controlled design. Here we show that a memory trace acquired by statistical learning is resistant to inference as well as to forgetting after one year. Participants performed a statistical learning task and were retested one year later without further practice. The acquired statistical knowledge was resistant to interference, since after one year, participants showed similar memory performance on the previously practiced statistical structure after being tested with a new statistical structure. These results could be key to understand the stability of long-term statistical knowledge.

Overlearning hyperstabilizes a skill by rapidly making neurochemical processing inhibitory-dominant

Overlearning refers to the continued training of a skill after performance improvement has plateaued. Whether overlearning is beneficial is a question in our daily lives that has never been clearly answered. Here, we report a new important role: Overlearning abruptly changes neurochemical processing to hyper-stabilize and protect trained perceptual learning from subsequent new learning. Usually, learning immediately after training is so unstable that it can be disrupted by subsequent new learning, unless waiting for passive stabilization, which takes hours. However, overlearning so rapidly and strongly stabilizes the learning state that it not only becomes resilient against, but disrupts, subsequent new learning. Such hyper-stabilization is associated with an abrupt shift from glutamate-dominant excitatory to gamma-aminobutyric-acid-dominant inhibitory processing in early visual areas. Hyper-stabilization contrasts with passive and slower stabilization, which is associated with a mere reduction of an excitatory dominance to baseline levels. Utilizing hyper-stabilization may lead to efficient learning paradigms.