Views from spike train

The billions of brain cells known as neurons that are organised into networks responsible for perception, the control of movement, memory, emotion, intelligence and consciousness, transmit messages in the form of nerve impulses. These are brief electrical pulses of constant amplitude - spikes - that can be conducted along the neuron at a rate of up to a few hundred per second. How the brain decodes the sequences of highly stereotyped spikes conveyed by sensory nerves to produce our perceptions of the world remains one of the most intriguing questions in the study of the nervous system and is the subject of this book.

The presence of a sensory stimulus can be signalled by a neuron either by a change in its firing rate, that is, in the number of nerve impulses generated in a given period of time, or by the timing of the individual impulses, which may be related to temporal variations in the stimulus. Beginning with the work of Edgar (later Lord) Adrian in Cambridge in the 1920s, most experiments in which electrical recordings have been made from sensory neurons have counted the spikes that follow presentation of a stimulus to assess how information about the outside world is transmitted within the brain. For example, neurons in the auditory part of the brain may show an increase in spike discharge when stimulated with sounds of increasing intensity; one would therefore conclude that this variable is signalled or encoded by changes in spike frequency or firing rate. But to understand how much information is contained in these trains of spikes and how this is used to reconstruct and make judgements about the world requires a much more rigorous approach. This was recognised in the 1950s and particularly by Horace Barlow, who was Adrian's student. With the advent of computational approaches to the study of the brain more quantitative attempts have now been made to describe and interpret the responses of neurons and especially those involved in audition and vision.

This book provides a modern appraisal of how sensory information is encoded and represented in the brain. Essentially the authors seek to define the meaning underlying patterns of nerve impulses by asking whether, given a particular spike sequence, it is possible to determine the nature of the stimulus. To achieve this, various mathematical approaches derived from information theory and statistical decision theory are applied to data obtained from studies carried out in species ranging from flies to monkeys. The authors analyse the responses of neurons to natural, dynamic stimuli and compare the reliability with which events can be detected and discriminated in relation to the behavioural performance of the whole animal. They show that neurons can be highly efficient, transmitting information with an accuracy and fidelity that is close to the physical limit imposed by noise in the system.

Spikes deals primarily with the information content of the spike sequences produced by single neurons. At first sight, this emphasis may seem surprising given that a sensory stimulus will potentially evoke a response from millions of neurons that are grouped together in particular regions of the brain. But if all the neurons respond in the same way then averaging the activity across an ensemble of neurons will not necessarily add any additional information and could degrade the message conveyed by the timing of the spike train. By adopting the rigorous analysis advocated by this book, it becomes apparent that behavioural decisions could potentially be made on the basis of individual spikes. On the other hand, there is also plenty of evidence to suggest that behaviourally significant information is contained in the global activity of populations of neurons with related but not identical functional properties. Moreover, recent studies have shown that sensory events can be specified with greater resolution when subsets of neurons fire in synchrony. It remains the job of the neurobiologist to design experiments that will demonstrate the relative capabilities of these different coding schemes and their relevance for behaviour.

Spikes is part of a series of monographs on computational neuroscience and is therefore aimed primarily at those with an interest in applying mathematical tools to analyse communication within the brain. Much of the text is used to explain and justify the quantitative methods needed to analyse spike trains in an attempt to convince the reader of the merits and shortcomings of each step. Most of the mathematical details needed to derive the equations given in the text are wisely confined to an appendix. Consequently, the text, which is written in a most engaging and lucid style, flows much more smoothly than it would otherwise do. At the same time, readers will need a reasonably good mathematical background to gain much more than just a few sound bites from this book.

As the authors point out, trying to unravel the meaning behind the sequences of impulses generated by cells in the brain is similar to learning a new language, although the dictionary currently available is far from complete. This authoritative and thoughtfully presented book represents a major step in defining the tools needed to decipher this language. Besides being of interest to those who approach the study of the brain from a theoretical standpoint, Spikes should certainly prove to be of practical value to neuroscientists engaged in attempts to measure and understand how our brains form a representation of what is going on in the world.

Andrew King is senior research fellow in physiology, University of Oxford.