Book review |
THE PHYSIOLOGY OF TRUTH
Jean-Pierre Changeux
2004. Cambridge, MA and London: Harvard University Press
Price £29.95.
ISBN 0-674-01283-6
University of Oxford, UK
Changeux on the promise of neuroscience
|
The thesis of The Physiology of Truth is that our brains have evolved to provide representations of the world and to make judgements about it that are useful for action. Some of these actions take place in social contexts and involve planning many steps ahead. In order to operate in this realm, the brain has needed to develop an understanding of social systems, strategies to achieve goals, and useful heuristics for guiding behaviour, including social behaviour. These heuristics lead Changeux to consider the biological background to morality and how plans to achieve long-term goals are produced. In seeking to understand how animals, including humans, work, Jean-Pierre Changeux is attempting to develop a scientific understanding of how and why we behave as we do.
Changeux draws on his own research—spanning the area from acetylcholine receptors to models of planning, how plans can be changed and the brain processes that may be related to consciousness—as well as on that of others. He appreciates the need to integrate knowledge derived from the spectrum of neuroscience disciplines to produce an explanation of how the brain functions. He realizes that, to understand how the brain performs its computations, it is important to know how single neurones are responding in particular situations, for it is single neurones that are the building blocks of computation by the brain. Single neurones are the information-processing elements of the brain; they act as non-linear computational devices and exchange information among themselves. It is only by analysing the spiking activity exchanged among neurones that one can understand how representations of information are provided by neural activity. It is then possible with modern computational neuroscience to take the responses of large groups of neurones and to understand their collective computational properties, as demonstrated in a series of interesting contributions by theoretical physicists and mathematicians (Amari, 1982
; Hopfield, 1982; Amit, 1989; Hertz et al. 1991; Rolls and Deco, 2002). Jean-Pierre Changeux has worked in many of these different areas.
Although Professor Changeux has combined information from many of these areas, one source that he might perhaps have drawn on more heavily is the activity of single neurones in the tasks that he discusses. This is an essential area of neuroscience in terms of understanding how the brain works, and it is now possible to model the activity of single neurones as they relate to, for example, changes in decision making as the required behavioural strategy changes. For example, Deco and Rolls (2003) produced a model of the prefrontal cortex that shows how the sets of neurones that map stimuli to actions can be switched using a top-down biased competition attentional modulation. The source of the modulation is a short-term memory that maintains a representation of the current mapping rule active. Deco and Rolls (2004) went on to consider how the controlling short-term memory attractor network could be switched as a result of changes in the rewards and punishments received.
One aspect of these particular models is that they start with the biophysics of single neurones (including the conductances that are opened at each synapse as a result of the inputs received by a neurone, and the time constants of the synapses and the membranes) to produce at this integrate-and-fire neuronal level a model of the dynamics of the operation of networks of large numbers of neurones. It is at this level that one makes direct contact with the activity of single neurones recorded neurophysiologically, for it is single-neurone spiking that is made explicit in these models.
The same integrate-and-fire models (Deco and Rolls, 2003, 2004) can be used to make predictions about what would be obtained at the next level up of explanation. For example, it is possible to integrate over the synaptic currents in the integrate-and-fire model of a particular brain area and to predict the functional magnetic resonance imaging (fMRI) BOLD (blood oxygenation level-dependent) signal that would result, given that it is ion flows across the neurones that require energy and thus lead to the BOLD signals that are measured using fMRI. An example of the way in which fMRI signals can be predicted from integrate-and-fire neuronal network models is described by Deco et al. (2004). We showed that differences between the dorsal and ventral parts of the prefrontal cortex during spatial and non-spatial aspects of a memory task could be related to different levels of neurophysiological inhibition in these areas.
The same type of integrate-and-fire neuronal network model can also be used to make predictions about the effects of brain damage. For example, Deco and Rolls (2002) showed, using a model of attention, how object-based visual neglect could arise as a consequence of interactions between graded increasing damage from right to left across a representation of visual space produced, for example, by a parietal cortex lesion and local lateral inhibition. These were the two essential components of an account of why patients with right parietal cortex damage might not see the left half of each of a series of objects spread out horizontally in the visual field.
Thus, part of the power of modern computational neuroscience is that it can help in producing models of behaviour that span from the biophysics of single neurones, the spiking neuronal activity of each neurone and the interactions of large populations of such neurones, to measures at a much higher level, for example of regional cerebral blood flow changes or the effects of brain damage. Jean-Pierre Changeux is interested in exactly this approach to understanding how the brain functions.
Changeux also considers some of the biological underpinnings of morality. Advances in this area have been made by understanding how animals, including humans, interact socially and have developed particular heuristics that are useful in optimizing social interactions. These approaches have led to the application of game theory to help understand some aspects of the way in which animals, including humans, may behave socially (Ridley, 1996). The modern field of neuroeconomics is also attempting to understand human decision-making, by taking into account not only the magnitudes of rewards, but also the probability that they will be obtained (Glimcher, 2004).
Thus, with great courage Jean-Pierre Changeux raises a very broad range of issues. There is much to be done before all these issues raised are understood satisfactorily, but the point that modern neuroscience can make very important contributions to understanding these major issues is cogently made.
References
Amari S. Competitive and cooperative aspects in dynamics of neural excitation and self-organisation. In: Amari S, Arbib MA. editors. Competition and cooperation in neural networks. Berlin: Springer; 1982. pp. 1–28.
Amit DJ. Modelling brain function. Cambridge University Press: New York; 1989.
Deco G, Rolls ET. Object-based visual neglect: a computational hypothesis. Eur J Neurosci 2002; 16: 1994–2000.[CrossRef][Web of Science][Medline]
Deco G, Rolls ET. Attention and working memory: a dynamical model of neuronal activity in the prefrontal cortex. Eur J Neurosci 2003; 18: 2374–90.[CrossRef][Web of Science][Medline]
Deco G, Rolls ET. A neurodynamical cortical model of visual attention and invariant object recognition. Vision Res 2004; 44: 621–44.[CrossRef][Web of Science][Medline]
Deco G, Rolls ET, Horwitz B. ‘What’ and ‘where’ in visual working memory: a computational neurodynamical perspective for integrating fMRI and single-neuron data. J Cogn Neurosci 2004; 16: 683–701.[CrossRef][Web of Science][Medline]
Glimcher P. Decisions, uncertainty, and the brain. Cambridge, MA: MIT Press; 2004.
Hertz JA, Krogh A, Palmer RG. Introduction to the theory of neural computation. Wokingham, UK: Addison-Wesley; 1991.
Hopfield JJ. Neural networks and physical systems with emergent collective computational abilities. Proc Natl Acad Sci USA 1982; 79: 2554–8.
Ridley M. The origins of virtue. London: Viking; 1996.
Rolls ET, Deco G. Computational neuroscience of vision. Oxford: Oxford University Press; 2002.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||