HFSP WORKSHOP II
VISION AND MOVEMENT IN THE CEREBRAL
CORTEX
A summary by science writer, Jennifer Altman
Putting together the pieces of vision
At the second Human Frontier workshop, an international group of scientists discussed how the brain processes the visual information used for perception and movement.
When we watch a moving ball, its colour and shape always stay together. Yet we should be surprised by this visual unity for investigations over the past 30 years have shown that this is not at all how the brain processes visual information. From the retina onwards, different aspects of our visual world are handled by distinct groups of nerve cells (neurons), often located in different parts of the brain. In April 1996, an international group of scientists met in Strasbourg to discuss the latest research on this parcellation and to tackle the much more challenging question of how we come to see the world as a whole.
In the 1980s, different parts of the cerebral cortex in monkeys were shown to deal with the questions "what is that object?" and "where is it?". "Where" takes the upper path, known as the dorsal stream, to the parietal cortex, located just above the ear, whereas "what" follows the lower road, or ventral stream, to the temporal cortex, lying under the ear. Brain imaging studies, described by Marc Jeannerod (INSERM, Bron, France) and Leslie Ungerleider (National Institute of Mental Health, Bethesda), have confirmed this separation in the human brain.
But this is far from the end of the story - over the past 15 years, nearly 30 areas (groups of neurons) in the cortex with different roles in processing visual information have been identified and the puzzle has been to work out what they do and how they work together. According to Jeannerod, recent thinking has put more emphasis on how the two streams regulate the behaviour of the animal. The dorsal stream, which leads to areas in the frontal cortex that control movement, deals with visually guided actions, such as reaching out and picking up a ball or a banana. In contrast, the ventral stream is concerned with perception of the object: identification and recognition. The two streams do talk to each other - monkeys respond differently to a banana and a snake - but how they exchange information is largely unknown.
Reaching and grasping : Recent research has shown that the dorsal stream is subdivided into several pathways with different functions: one regulates reaching out to contact an object, while another determines how the hand grasps shapes as diverse as ball and banana. Yet another deals with the movements of the eyes as they follow a moving object. Each pathway links together several areas with related functions but even in a single area there may be neurons responding to different aspects of the action to be accomplished. For example, Francesco Lacquaniti (ICRS, Rome, Italy) described neurons in one area that signal either the horizontal or vertical extent or the distance the arm needs to move to reach an object. Similarly for manipulation, Hideo Sakata (Nihon University, Japan) has found neurons in part of the parietal cortex that respond to the sight of the object, others that signal the hand movement needed to grasp it and yet others that combine vision and movement.
Keeping track of a moving object means moving the eyes and often the head. Both Richard Andersen (California Institute of Technology) and Michael Goldberg (National Eye Institute, Bethesda) have been examining the role of neurons in one part of the parietal lobe during such eye movements but they reported rather different conclusions: Andersen considers most neurons in this area signal the animal?s intention to make an eye movement, whereas Goldberg thinks they predict the direction in which the eyes must look. This disagreement highlights the difficulties inherent in interpreting experiments that require monkeys to learn to make complex sequences of movements.
Not only the cortex is involved in determining ?where?. Klaus-Peter Hoffmann (Ruhr University, Bochum, Germany) described how subcortical areas that regulate the slow movements of the eyes as they follow an object receive inputs from specific motion sensitive neurons in the cortex. He also showed how another subcortical structure, the superior colliculus, which is already known to participate in the control of eye movements, has a parallel role in visually guided arm movements.
Object recognition : Work on the ventral stream has focused on the way the brain analyses the appearance of objects. Different groups of neurons in the cortex respond to various components of the object, such as shape, texture and colour. The signals from these various areas converge in the temporal cortex, where neurons respond to moderately complex combinations of attributes. So far, single neurons that signal the whole of a complicated object have not been found. Rather, recognising a specific object or face seems to work by combining several ?building block? responses. As Keiji Tanaka (RIKEN, Japan) described, neurons responding to similar features of objects or different views of the same object are clustered together. Some neurons have surprisingly specific requirements: David Perrett (University of St Andrews, UK) showed one that responded to several views of the same face, providing the eyes were looking in a particular direction.
Populations and networks : This mix-and-match system provides the flexibility needed for recognising unfamiliar objects. It is one variant of an important concept, termed population coding, that has emerged from studying neurons in the motor cortex, the region where the instructions to the muscles are formulated. Here, the direction of a particular movement is not determined by specific neurons. Instead, it is represented by the combined strengths of the signals of a large number of neurons. Apostolos Georgopoulos (Brain Sciences Center, Minneapolis) thinks it is likely that many areas of the brain use some form of this coding principle.
Another emerging concept, stressed by Roberto Caminiti (University of Rome, Italy), is that the areas that making up a pathway operate as a network. Instead of signals being passed from one area to the next in a linear hierarchy, all the areas talk to each other, so that incoming information is modified by experience and memory. One example, reported by Nikos Logothetis, is the interpretation of ambiguous pictures, such as the figure that appears as a vase or two faces. Resolving such conflicts seems to involve many stages in the ventral pathway, rather than a single area.
Working memory, which holds information temporarily while a problem is being solved, also seems to depend on a network of areas. The dorsal and ventral streams send information to different parts of the prefrontal cortex known to be centres for working memory. Both Ungerleider and Patricia Goldman-Rakic (Yale University School of Medicine) are now finding that working memory also involves parts of the temporal and parietal lobes. Goldman-Rakic considers there may be parallel working memory networks for spatial analysis, object recognition and even for language.
Binding : How the brain puts all the pieces of information back together is still unsolved. One thing seems clear - there is no single area in the brain dedicated to interpreting the visual world. Wolf Singer (Max Planck Institute for Brain Research, Germany) proposed that the responses of the many neurons activated by the attributes of a banana are synchronised so that they work as an assembly that may be distributed across several areas. The concepts of population codes and networks could provide a basis for understanding how such synchronised assemblies may form and operate.
Return to Table of Contents