The eyes have it, even if the message is a blur

If I were to choose a single book on vision with which to be marooned on a desert island, it would have to be The Visual Neurosciences . In between paddles in the ocean and experiments with sandcastles, the 114 chapters of this enormous book would be invaluable in providing mental stimulation to while away the lonely hours.

Modern visual neuroscience is an interdisciplinary enterprise, and the study of vision is in many ways a model for the way knowledge from different fields can be brought together to study the brain. Editors Leo Chalupa and John Werner have managed to assemble a remarkable collection of contributions from the world's leading vision scientists.

The work comes as close as possible to encapsulating the state of knowledge in the visual neurosciences, a dynamic and diverse field spanning anatomy, physiology and psychology and sharing borders with philosophy, computer science and molecular biology, among other disciplines. It is divided into thirteen sections, typically comprising five to twelve chapters. The contributors include many of the leading names in world neuroscience. Each chapter is a summary of current evidence and thinking in the author's area of expertise.

Volume one takes us through the first seven sections. It is concerned primarily with the lower levels of visual processing, proceeding from "Historical foundations" and "Developmental processes" through "Retinal mechanisms and processes", "Organisation of the visual pathways" and "Subcortical processing", culminating in sections on "Processing in the primary visual cortex" and "Detection and sampling". Volume two is concerned almost exclusively with higher-level, cortical vision. Its organisation in part reflects the apparent modularity of cortical vision, beginning with sections on "Brightness and colour", "Form, shape and object recognition" and "Motion, depth and spatial relations". These are followed by sections on "Eye movements", "Attention and cognition" and "Theoretical and computational perspectives".

Section one contains chapters on the historical foundations from two of the elder statesmen of visual neuroscience, Mitch Glickstein and Horace Barlow.

Glickstein outlines the discoveries that allowed understanding to progress from image formation by the eye, through the role and operation of the photoreceptors, to the connections between eye and brain. Barlow concentrates on the role of single-cell electrophysiology in the development of visual neuroscience. As well as giving an overview of past research, in particular the use of the statistical approach of signal detection theory in single-cell studies, he speculates that much will be learnt in the future by investigating more fully the complex computational capabilities of pyramidal cells and the multiple timescales at which they appear to operate.

Section two, on developmental processes, begins with descriptions of the biochemical and electrical underpinnings of retinal development. Ganglion cell axons from the two retinas meet at the X-shaped optic chiasm and seemingly must choose to which side of the brain to project. This simple choice point is an ideal context in which to study the molecular mechanisms of axon guidance in the developing nervous system. The ultimate goal of these axons is the visual thalamus, which develops as a layered structure with inputs from the two eyes segregated into different laminae. This segregation is a competitive process dependent on spontaneous activity from the prenatal retina. Eye-specific segregation is also clearly evident at the entry point to primary visual cortex, layer IV, which is organised into alternating columns of ocular dominance. The formation of these columns, as well as the development of the orientation selectivity characteristic of the primary visual cortex, depends in a complex way on patterns of neuronal activity within the developing visual pathway.

Section three, on retinal mechanisms and processes, describes the function of the many classes of cell in the retina. Visual processing begins when light is transduced into electrical signals in the rod and cone photoreceptors. This transduction process is now understood in great detail, revealing the sophisticated operations that mediate the transfer of visual information even at the earliest stages of processing. A simple hierarchical view of the retina is that the photoreceptors feed into bipolar cells and thence into the retinal ganglion cells, whose axons send signals down the optic nerve and towards the brain. However, such a description neglects much of the circuitry involved in retinal processing of the visual signal. Horizontal cells form lateral interconnections between photoreceptors, while other horizontal connections within the retina are mediated by a wide variety of amacrine cells. Such is their significance to visual processing that both the horizontal and the morphologically marvellous amacrine cells are the subject of dedicated chapters.

Section four gives an overview of the organisation of the visual pathways.

The lower levels of the primate visual system contain (at least) three parallel pathways originating in the magno, parvo and koniocellular retinal ganglion cells. These pathways each have distinct patterns of projection into the visual cortex and possibly subserve different functions. A functional distinction between parallel pathways appears clearer at the higher levels of the visual pathway beyond the primary visual cortex.

There, processing is characterised by a dorsal stream specialised for the analysis of motion and visual space and a parallel ventral stream specialised for the recognition of objects.

Section five surveys the subcortical pathways connecting the retina and the cortex, such as the thalamus and the pulvinar, and those mediating the response of the pupil. Again, there is emphasis on the need to look beyond the simple, hierarchical view of visual processing, this time in considering the flow of information from thalamus to the primary visual cortex to extra-striate cortex. There are a vast number of feedback connections from the primary visual cortex to the visual thalamus and from extra-striate areas back to the primary visual cortex. Although there is clear evidence that these feedback loops play a critical role in visual processing at the thalamic and cortical levels, we still have surprisingly little idea what it might be.

Section six focuses on processing in the primary visual cortex. Cells in the primary visual cortex typically have small receptive fields. That is, they are only excited by stimulation within a highly localised region of the visual field. However, their response properties are often much more complex than those of the thalamic neurons that provide their input. For example, although the receptive fields of cells in the primary visual cortex are small, the response elicited by stimulation of the receptive field can be modulated by the structure of the surrounding image. Such contextual modulation is a characteristic not only of cortical visual processing but also of our perceptual experience. Indeed, the primary visual cortex is the earliest stage of the visual hierarchy at which correlates between neural processing and visual perception have been reported. The primary visual cortex is also the site at which signals from the two eyes are first combined. It is consequently where the processing of stereoscopic depth begins. However, consideration of the response properties of cells in the primary visual cortex demonstrates that they carry out only the first in a chain of processes necessary to perceive depth.

Section seven, on detection and sampling, begins with three largely theoretical contributions on the formation of the retinal image and the statistics of detecting visual signals in the inevitable presence of neural noise. These are followed by three chapters describing vision adapted to different levels of ambient illumination. In night vision, the dark-adapted eye relies on the activity of rods to detect what little light is available. Full daylight saturates the rods but provides enough energy to stimulate the three classes of cones whose relative activity allows the extraction of chromatic information. At intermediate twilight levels, the visual world comes to us through an interaction of rod and cone signals.

Section eight deals with brightness and colour. Our colour vision allows us to recognise objects on the basis of their surface reflectance despite huge variations in the ambient illumination. The fact that we are almost unaware of the effect of illumination on the spectral composition of the light reaching us from an object is a manifestation of the mechanisms of colour constancy in our visual systems. These retinal and cortical mechanisms take into consideration both spatial and temporal context in computing colour.

Section nine covers form, shape and object recognition. We are able to recognise objects on the basis of their shape despite variations in viewpoint and viewing distance. Our visual systems must parse the image into meaningful segments and use shading and texture cues to infer three-dimensional shape. Single-cell studies on monkeys and human brain imaging demonstrate that this is the work of the ventral processing stream.

Section ten, on the processing of motion, depth and spatial relations, begins with a chapter on motion cues in insect vision and navigation, reflecting the fact that motion processing is one area of vision where what has been learnt about insect visual processing has proved relevant in our understanding of human vision. Image motion occurs whenever there is relative motion between our eyes and the world around us. It is caused not only by motion in our surroundings but principally by our own motion through the world. Our visual systems must disentangle these contributions to recover a stable representation of our environment. This section also deals with the perception of depth, a topic first touched on in section six. There it was argued that processing in the primary visual cortex is insufficient to account for our perceptual awareness of depth. In this section, the links between processing and perception are again a focus. One chapter details the close correspondence between activity in the middle temporal visual area and the perception of motion. Another describes the use of binocular rivalry - the alternating perception set up when dissimilar images are presented to the two eyes - to distinguish between stages of the visual hierarchy that are passively processing the retinal image and areas of the brain that are involved in perception.

Section 11 deals with eye movements. In normal vision, our eyes are almost constantly in motion, drifting smoothly or making saccadic jumps from one part of the visual scene to another. Our visual input is all but wiped out during saccades as the image is blurred across our rapidly moving retinas.

Remarkably, though, our eye movements rarely impinge upon our conscious perception. This observation leads to several important questions. How is our perception of the retinal image suppressed during saccadic eye movements? Why, if our perception of the retinal image is suppressed several times per second, are there not corresponding gaps in our perceptual experience? And what is it that forms the content of our visual awareness during saccades?

Section 12, on attention and cognition, looks at the neural basis of visual attention. We all know subjectively what it is to attend to something, but what is happening in our brains while we are doing so still remains largely mysterious. I found this section tantalisingly brief. It contains only five chapters and the one that I was most looking forward to, John Maunsell's discussion of the role of attention in the visual cerebral cortex, was only six pages long. This is the one section of the book that I find inadequate in its coverage of an increasingly important topic.

Section 13, "Theoretical and computational perspectives", is a rather eclectic collection of chapters, many of which might have sat more comfortably in earlier sections. The final chapter by Christof Koch and the late Francis Crick outlines their ideas about the neural basis of visual consciousness.

To sum up, The Visual Neurosciences is a major work that should be found on the bookshelf of every vision scientist. It will be a valuable reference for graduate students and other researchers in the field.

Colin W. G. Clifford is senior lecturer in psychology, University of Sydney, Australia.