Perception, learning and memory

Perception, learning and memory are interconnected processes controlled by the coordinated activity of molecules, synapses, cells and neural networks within the brain.  Although we know much about the activity of individual neurons and synapses, we know far less about how these components interact. Neuroscience techniques must evolve to realize a new era of multidisciplinary research studying networks of interacting elements.

The human brain is a highly complex organ shaped by hundreds of millions of years of evolution. It has evolved to detect meaningful patterns, to learn, memorize and recall them, and to adapt. Our neural networks can produce and decode communication signals, extract and process useful features from the environment, and produce vital innate behaviours such as eating, fleeing and mating. Amazingly, this specialized structure selfassembles, growing from one cell to tens of billions, and each developing brain incorporates both hidden biases shaped through natural selection, and the means with which to sculpt itself throughout its lifetime as the individual encounters new experiences and sensations.


Our brain contains billions of neurons, which are specialized cells that process and transfer information, and are arranged into complex cellular circuits. These cells communicate via synapses, which are junctions that allow the transfer of chemical or electrical information from one neuron to the next (Fig. 1).

Neurons are the most diverse cell type in the body. They are usually polarized with specialized projections for receiving (dendrites) and relaying (axons) information (Fig. 2). Sensory neurons convert external stimuli, such as light, sound or pressure, into electrical signals, whereas motor neurons use electrical signals to control muscles. A third, more abundant, type of neuron lie between these inputs and outputs.

Non-neuronal cells, called glia, play fundamental roles in the development, support and plasticity of neural circuits; however, neurons and their synapses remain the focus of learning and memory research. Changes in neuronal activity and synaptic strength are thought to underpin learning and memory. Moreover, neuronal loss and synaptic malfunctioning have been implicated in various neurological disorders that involve learning and memory deficits.


Despite intense efforts to understand perception, learning and memory, there are still huge gaps in our knowledge. We have yet to pinpoint the neuronal mechanisms that underlie perception. Millions of neurons in structurally diverse networks activate for us to perceive even the simplest of objects, and teasing apart the neural circuits that are responsible is no small achievement.

Perception and memory are intimately interlinked — perceiving an object would be meaningless without the ability to recall and link it to corresponding memories. Although perception, memory formation and recall are likely to rely on interlinked mechanisms and substrates, we have yet to understand them fully, or to decipher the effects of sleep, attention and other ill-understood processes on learning and memory.

We do know that memory is a spatially and temporally dynamic process. As memories are stored and consolidated, they are shifted from one part of the brain to another in a process that can take weeks and appears to be dependent on brain activity during certain phases of sleep1. Memory-related proteins, synapses, neurons and neural networks are also dynamic. Neurons die off as part of normal ageing, yet for the most part we notice no change. Proteins are continually recycled and replaced, and new proteins are required for learning and memory to occur2. So, how can some memories remain stable when so many of the underlying components are constantly changing?


The study of perception, learning and memory offers many challenges and research opportunities. Our technical toolbox means it is now possible to catalogue and describe the constituents of the brain and its neural circuits — an essential step towards understanding the brain. All that is needed is the time and optimized methods to help handle huge data sets.

The development of high-resolution serial electron microscopy, super-resolution light microscopy3,4 and multicoloured genetic tools for neuronal labelling5, and the increased affordability of immense computing power, make it possible to imagine a day when the connection matrix of a small-to-medium-sized brain (perhaps that of a fly or a mouse) will be known with a reasonable degree of accuracy. This endeavour requires the ability to handle enormous data sets, and a multi-disciplinary approach incorporating molecular biology, genetics, electrophysiology, imaging, electronics, nanotechnology, mathematics, computer science and nonlinear dynamics. It will breed a new type of cooperation between areas of science that have often worked separately.


A major challenge for modern neuroscience is to explain perception and behaviour in terms of neural activity. Given the size of the brain, the number of neurons and the distributed nature of neural activity6, it is increasingly clear that traditional methods will yield limited results. Patch clamping, for example, can record the activity of single cells at high resolution, but tells us nothing of how these cells contribute to larger circuits. Functional magnetic-resonance imaging offers a broad view of brain activity on a large scale, but lacks the resolution to reveal the activity of individual neurons (Fig. 3).

Much is known of the functioning of individual neurons and synapses, but much less about their coordinated action in ensembles of millions. The brain derives its magic from coordinated activity on the large scale and high degrees of specialization on the small scale7.

>> Much is known of the functioning of individual neurons and synapses, but much less about their coordinated action in ensembles of millions.

Networks, neurons and molecular constituents need to be studied in combination rather than in isolation, and experimental techniques traditionally used to study individual elements need to evolve towards this. One new approach involves light-activated genetic switches that control the activity of specific, discrete neuronal populations8,9. This technique — ‘optogenetics’ — is already bearing fruit and it is thought that such studies will help reveal cell function within the context of neural circuits.

Neural activity needs to be sampled at an intermediate scale: that of networks of interacting elements. Rather than studying a handful of cells in a handful of animals, studies should focus on the population level, with high-sampling density and mobile animals. This will be technically challenging, and will rely on major developments in the fields of optics, microelectronics, nanoelectronics and computer science.

The rewards will be great. Deciphering the neural basis of perception, learning and memory is a fundamental part of understanding how the brain functions in health, ageing and disease. Teasing apart the contributory mechanisms might offer us the chance to influence and improve these most human of skills.

The dynamic and coordinated behaviour of neurons in the brain can be detected in brain oscillations that occur at a variety of frequencies (for example, 2–200 Hz). a recent study by researchers at the Max Planck Institute for Brain research found that when memory-related neurons in the brain fire synchronously with brain waves at the theta frequency (2–8 Hz) during learning, the resulting memories are stronger than if this synchronization does not occur (Rutishauser, U. et al., Nature 464, 903–907, 2010).

Wierzynski, C. M.
State-dependent spike-timing relationships between hippocampal and prefrontal circuits during sleep.
Neuron 61, 587–596 (2009).
Sutton, M. A. & Schuman, E. M.
Local dendritic protein synthesis and plasticity.
Cell 127, 49–57 (2006).
Nägerl, U. V. et al.
Live-cell imaging of dendritic spines by STED microscopy.
Proc. Natl Acad. Sci. USA 105, 18982–18987 (2008).
Westphal, V. et al.
Video-rate far-field optical nanoscopy dissects synaptic vesicle movement.
Science 320, 246–249 (2008).
Lichtman, J. W., Livet, J. & Sanes, J. R.
A technicolour approach to the connectome.
Nature Rev. Neurosci. 9, 417–422 (2008).
Singer, W.
Distributed processing and temporal codes in neuronal networks.
Cogn. Neurodyn. 3, 189–196 (2009).
Laurent, G.
Olfactory network dynamics and the coding of multidimensional signals.
Nature Rev. Neurosci. 3, 884–895 (2002).
Nagel, G. et al.
Channelrhodopsins: directly light-gated cation channels.
Biochem. Soc. Trans. 33, 863–866 (2005).
Nagel, G. et al.
Channelrhodopsin-2, a directly light-gated cation-selective membrane channel.
Proc. Natl Acad. Sci. USA 100, 13940–13945 (2003).
Rutishauser, U.
Human memory strength is predicted by theta-frequency phase-locking of single neurons.
Nature 464, 903–907 (2010).

Go to Editor View