In this special issue of Nature Neuroscience, we feature an assortment of reviews and perspectives that explore the topic of learning and memory.
Focus on Learning and Memory
Learning new information and skills, storing memories of this knowledge, and retrieving, modifying, or forgetting these memories over time are critical for flexibly responding to a changing environment. How these processes occur has fascinated philosophers, psychologists, and neuroscientists for generations, and continues to inspire research encompassing diverse approaches. In our October 2019 issue, Nature Neuroscience presents a collection of reviews and perspectives that reflects the breadth and vibrancy of this field.
In this Review, Miller and Sahay discuss how adult-born neurons recruit inhibitory microcircuits to support hippocampal memory indexing and pattern separation.
Memory retrieval involves interactions between internal or external cues and stored engrams. Identification of engrams in mice permits examination of these interactions at the level of neural ensembles. This review highlights emerging findings.
In this Review, Likhtik and Johansen discuss how modern neuroscience techniques applied to the study of emotional learning reveal new principles for how neuromodulatory systems regulate distributed brain circuits and flexibly adjust behaviour.
While we sleep, the brain replays memories of our experiences during the day. In this review, Klinzing et al. provide a concise overview of how the sleeping brain transforms and builds persisting memories through this process.
The authors review the most recent measurement and manipulation approaches that enable links between synaptic plasticity and learning to be examined, and they propose potential future approaches to tackle this endeavor.
When crossing the street, you can ignore the color of oncoming cars, but for hailing a taxi color is important. How do we learn what to represent neurally for each task? Here, Niv summarizes a decade of work on representation learning in the brain.
This paper first reviews the work on brain-machine interfaces (BMIs) for restoring lost motor function and then provides a perspective on how BMIs could extend to the new frontier of restoring lost emotional function in neuropsychiatric disorders.
A deep network is best understood in terms of components used to design it—objective functions, architecture and learning rules—rather than unit-by-unit computation. Richards et al. argue that this inspires fruitful approaches to systems neuroscience.
From the archive
The function of rapid eye movement (REM) sleep remains unclear. By examining how REM sleep affects synapses in the mouse cortex, the authors show that REM sleep is fundamental to brain development, learning and memory consolidation by selectively pruning and maintaining newly formed synapses via dendritic calcium spike-dependent mechanisms.
The authors address why the use of prior expectations might be compromised in autism, by using computational models and pupillometric markers of the neuromodulator noradrenaline. They show that by estimating the world to be more changeable than it really is, adults with autism have difficulty in learning what to expect.
PTSD symptom severity in combat veterans was associated with enhanced sensitivity to prediction errors and lower neural tracking of value and learning rate, providing evidence for neurocomputational contributions to trauma-related psychopathology.
Impaired perceptual learning in a mouse model of Fragile X syndrome is mediated by parvalbumin neuron dysfunction and is reversible
Goel et al found similar deficits in visual discrimination in humans and in a mouse model of FXS. In mice, a robust decrease in PV cell activity mediated this impairment, suggesting that manipulating inhibition may improve sensory processing in FXS.
Humans and other mammals are prodigious learners, partly because they also ‘learn how to learn’. Wang and colleagues present a new theory showing how learning to learn may arise from interactions between prefrontal cortex and the dopamine system.
Single-cell activity tracking reveals that orbitofrontal neurons acquire and maintain a long-term memory to guide behavioral adaptation
Namboodiri, Otis et al. reveal that orbitofrontal cortex acquires and maintains a long-term memory of cue–reward associations to guide multiple aspects of behavioral learning, and that it routes select information to a downstream learning center.
The DNA modification N6-methyl-2’-deoxyadenosine (m6dA) drives activity-induced gene expression and is required for fear extinction
Li et al. have discovered a necessary role for the DNA modification N6-methyldeoxyadenosine (m6dA) in regulating experience-dependent gene expression and the formation of fear extinction memory. These findings expand the scope of DNA modifications in the adult brain.
Pairing an odor conditioned stimulus (CS) with an unconditioned stimulus (US) induces memory formation. Vetere et al. replace the real CS and US with direct optogenetic stimulation of the brain and create a fully artificial odor memory in mice.
Memory formation depends on both synapse-specific modifications of synaptic strength and cell-specific increases in excitability
The authors discuss newly emerging evidence for the role of the transcription factor CREB in memory, including its role in modulating changes in excitability that are critical for neural assembly formation and linking of memories across time.
Prefrontal cortex can be flexibly engaged in many different tasks. Yang et al. trained an artificial neural network to solve 20 cognitive tasks. Functionally specialized modules and compositional representations emerged in the network after training.
By pairing cues with brief activation of dopamine neurons in absence of reward, the authors reveal elemental behaviors conditioned by dopamine, showing VTA underlies generation of incentive value and SNc supports conditioned movement invigoration.
As naive mice learn a stimulus–reward association, DA neuron activity first reflects the timing of reward-seeking actions relative to predictable stimuli & rewards. As actions are refined by learning, DA neuron activity can reflect prediction errors.
Learning to predict reward is thought to be driven by dopaminergic prediction errors, which reflect discrepancies between actual and expected value. Here the authors show that learning to predict neutral events is also driven by prediction errors and that such value-neutral associative learning is also likely mediated by dopaminergic error signals.
The authors show how predictive representations are useful for maximizing future reward, particularly in spatial domains. They develop a predictive-map model of hippocampal place cells and entorhinal grid cells that captures a wide variety of effects from human and rodent literature.
The authors develop a noninvasive stimulation protocol to restore neural synchronization patterns and improve working memory in older humans, contributing to groundwork for future drug-free therapeutics targeting age-related cognitive decline.
Learning is ubiquitous in everyday life, yet it is unclear how neurons change their activity together during learning. Golub and colleagues show that short-term learning relies on a fixed neural repertoire, which limits behavioral improvement.
Corticospinal cells of the motor cortex act as a direct link between the cortex and movement-generating circuits within the spinal cord. The authors demonstrate that the relationship between activity of these cells and movement changes with time and learning, indicating a flexible cortical output to drive movements.
Granule cells constitute half of the cells in the brain, yet their activity during behavior is largely uncharacterized. The authors report that granule cells encode multisensory representations that evolve with learning into a predictive motor signal. This activity may help the cerebellum implement a forward model for action.
Distinct learning-induced changes in stimulus selectivity and interactions of GABAergic interneuron classes in visual cortex
Khan et al. simultaneously measured activity from excitatory cells and three classes of inhibitory interneurons in visual cortex and show that learning differentially shapes the stimulus selectivity and interactions of multiple cell classes.
Using a deep learning approach to track user-defined body parts during various behaviors across multiple species, the authors show that their toolbox, called DeepLabCut, can achieve human accuracy with only a few hundred frames of training data.
Like humans, songbirds learn to communicate vocally early in life. Moore and Woolley taught birds the songs of a different species to identify how vocal experience and auditory tuning mechanisms create neural representations of communication sounds.
Effective learning is accompanied by high-dimensional and efficient representations of neural activity
Why do certain individuals learn faster than others, and how does the acquired information take shape within their brains? Tang and colleagues show that fast learners encode information in a particularly compact, efficient and space-saving manner.
Lacagnina et al. show that extinction training suppresses the associated hippocampal fear engram and generates a distinct extinction engram. Reactivation of extinction engram cells reduces fear, while reactivation of fear engram cells causes fear relapse.
A simple behavioral task identifies two qualitatively different groups within the general population, according to their speech-to-speech synchronization abilities. Group pertinence predicts brain function and anatomy, as well as word-learning performance.
PKCα integrates spatiotemporally distinct Ca2+ and autocrine BDNF signaling to facilitate synaptic plasticity
Through the development of novel PKC biosensors, the authors describe how PKCα, but not other classical isozymes, facilitates plasticity in dendritic regions through the integration of recent synaptic plasticity with current, local synaptic input.