Superior Index Go to the next: Chapter 2
Print Files: A4 Size.
In science, the nature of explanation is most often such that some empirical phenomenon is considered explained if it can be related to previously established fundamental principles. For example, the motion of a baseball is explained at the beginning of the 21st century in terms of the force imparted by the pitcher's arm, the resistance of the atmosphere, the air turbulence caused by the spin of the ball and, most of all, in terms of the force of gravity. But what about the situation in 1620 with regard to the explanation of the motion of planets? It is technically true to say that Newton "explained" planetary motion in terms of the fundamental principle of gravitational attraction, but "explained" has a subtly different sense here. The principle of gravitation, or at least its universality, was actually being established by Newton's explanation of the motion of planets. So in this case, Newton was not relating a single empirical phenomenon to previously established fundamental principles; he was relating several otherwise independent phenomena (Kepler's three laws of planetary motion) to a common underlying principle which thereby became established.
In other words, when there are already established fundamental principles in an area, an empirical phenomenon is considered to be explained if it can be related to these fundamental principles. But if there are no fundamental principles already established, such principles must first come to be established. The way this generally happens is that the fundamental principles are first proposed and then shown to be valid by showing that they explain many otherwise unrelated observations.
In essence, what this book is doing is proposing a basic fundamental scientific principle regarding consciousness and then seeking to establish this principle, by showing that it explains many observations.
The basic fundamental principle being proposed here is that consciousness is identical with certain spatiotemporal patterns in the electromagnetic field.
At this stage we can say nothing about the shape of these spatiotemporal electromagnetic patterns or the nature of the rules that connect such patterns with subjective experience. Those details will come later. What we need to do first is try and establish the truth of the basic fundamental principle that certain spatiotemporal electromagnetic patterns are identical with conscious experiences.
The analysis above about how fundamental principles are generally established in science would suggest that all we need do is come up with some measured electromagnetic field patterns that always covary with conscious experiences. If we can do this, we should have established our fundamental principle, by showing that it does tie together all the observations that have so far been made (and that there is reasonable hope that it will tie together observations yet to be made) about conscious experiences. An optimist, or indeed a historian of science, might well accept this proposal.
However, suppose we did find that certain spatiotemporal patterns in the electromagnetic field always covary with particular conscious experiences. What would be the actual effect of this on the minds of real-life scientists? There are basically two possible interpretations of such a finding:
The attitude of most physiologists towards the brain-generated electromagnetic patterns that can be measured by EEG recording has traditionally been option (b) - that they are meaningless epiphenomena, merely the smoke produced by the fire of brain activity. This view was strongly held by the charismatic neurophysiological pioneer Sir John Eccles, who was very influential in promulgating the idea that the most important thing for physiologists to study was activity in single cells and synapses. Oddly enough the other arm of Eccles' beliefs about the mind, which was essentially dualist, is now held by physiologists (who in general subscribe to the psychoneural identity theory) to be downright heretical - but his influence in the matter of the epiphenomenality of the EEG has remained strong.
On the other hand, the view in disciplines other than neurophysiology has traditionally been much more open to the idea that electromagnetic patterns have a direct ontological relationship with consciousness. In general this view is tenable (though never expressed openly) in university departments of physics and engineering and also in departments of psychiatry, probably because the human EEG was discovered by the psychiatrist Hans Berger.
The other group of scientists who one would have thought should have a professional interest in consciousness are psychologists. However oddly enough, psychologists in the century or so after William James became completely sidetracked by the Behaviorist paradigm into believing that consciousness didn't actually exist at all, or at least that it was not to be mentioned in polite academic company, and it is only relatively recently that the cognitive revolution has restored consciousness to the status of a fit subject for study in psychology departments. As a result of this strange situation, for some decades a knowledge of brain structure and function was not deemed necessary for the study of psychology. Perhaps because of their relatively recent conversion to the opposite viewpoint, cognitive psychologists today tend to be more inclined to the neurophysiological approach to the study of mind i.e. either to take for granted some version of the neural identity theory, or to be functionalists who see consciousness as a process rather than any form of substance.
So, reaction to a finding that there is a strong correlation between electromagnetic patterns and consciousness is likely to be split fairly cleanly along disciplinary lines (with of course, a few honorable exceptions). Some will accept the possibility that correlation in this case indicates identity, but rather more will naturally adopt the opposite viewpoint. Rather than plunge into the philosophical swamp on this issue, let us just flag it as a potential problem and set it aside. The first thing to determine is whether or not there do actually exist patterns in the electromagnetic field which carry all the necessary information for distinguishing, from the outside so to speak, one experience from another. In other words, the question is, do there exist electromagnetic patterns which reliably predict the existence of particular conscious experiences? If there do, then the philosophical question of identity may become worth discussing.
In order to answer the question of whether or not there actually exist the sort of correlations between conscious experience and electromagnetic patterns which we are looking for, we must first lay down some working definitions and also some ground rules for acceptability of experimental evidence.
The definition of the word consciousness has stopped in their tracks more academic discussions than you could shake a stick at. It would be possible to write a whole book on the subject of the definition of the word consciousness (and indeed in some senses this is exactly what the current book is about). But whatever our conclusions may turn out to be at the end of this exercise, clearly some attempt must be made at the outset to state as plainly as possible what it is that we are talking about.
Perhaps the obvious first step is to consult a dictionary for a definition of consciousness. The Pocket Oxford Dictionary at first appears to be admirably concise on the subject. It says that the noun "consciousness" means "awareness; person's conscious thoughts and feelings as a whole". Unfortunately this is impression of precision is rather spoiled when one looks up the adjective "aware" and finds that it means "conscious". Clearly there is no help to be had from this source, other than the information that the words conscious and aware are generally regarded as being synonymous.
Where to now? Since words are the stock in trade of philosophers, perhaps we should see what philosophers have to say about the definition of the word conscious. John Searle offers the following (Searle, 1993 [255]):
"Like most words, `consciousness' does not admit of a definition in terms of genus and differentia or necessary and sufficient conditions. Nonetheless, it is important to say exactly what we are talking about, because the phenomenon of consciousness that we are interested in needs to be distinguished from certain other phenomena such as attention, knowledge and self-consciousness. By `consciousness', I simply mean those subjective states of sentience or awareness that begin when one wakes up in the morning from a dreamless sleep and continue throughout the day until one goes to sleep at night or falls into a coma, or dies, or otherwise becomes, as one would say, `unconscious'."
This definition is fine in that it does basically tell us what we are talking about, but not terribly satisfactory in that it says nothing about how one might determine whether or not an entity other than ourselves is conscious. The classical Turing test (which basically defines an entity as intelligent if it is not possible to distinguish between it and another human in casual conversation) can clearly be passed by simple computer programs which no-one would accept as being conscious in the same sense as humans are, but would not be passed by great apes, to which most people would be inclined to grant the possession of at least simple consciousness. Unfortunately however, there appears to be no better test for the presence of another consciousness available at this time. My own preference is simply to sidestep the issue, by saying only that I know what conscious experience means in my own case and I infer from your behavior and the fact that you look roughly like me that it means something similar for you, dear reader (although I can never actually be sure of that).
One further refinement of the definition of consciousness which may be useful is provided by the psychiatrist Richard Maurice Bucke. In his classic description of mystical experience (Bucke, 1993 [45]), this author divides consciousness into three types or grades: simple consciousness, self consciousness and cosmic consciousness. Simple consciousness is possessed by humans and also by animals such as cats and dogs - it consists largely of sensory experiences and perceptions. Humans usually make the transition at some time in their childhood from simple consciousness to self consciousness, which is possessed by all normal adult persons. This consists of a state in which one is aware that one is an individual, a self. Cosmic consciousness is described by Bucke as a further state, which has been attained by only a few humans in the history of the race (although the author suggests that the number with this faculty is increasing over the centuries) and almost never appears before the age of about 35. It consists of an ongoing, direct awareness that one is not actually an individual after all, but merely part of an all-encompassing, immortal Self. The stereotyped mystical experience in which this is realised for the first time is generally called the unity experience. That cosmic consciousness is not merely a pathological state is evidenced by the major accomplishments in science and the arts of those whose writings suggest they experienced it.
The main part of this book is about simple consciousness, which consists of sensory experiences. So unless otherwise specified, the word consciousness will hereafter be used to mean simple consciousness. However it will be seen in the last chapter that the theory expounded here about the nature of consciousness just as easily incorporates or provides an explanation for cosmic consciousness.
In thinking about simple consciousness, is convenient to distinguish between general states of consciousness and specific contents of consciousness. The term "states of consciousness" is usually taken to refer to the states of waking, sleeping and dreaming. The absolute minimum that must be done by any theory of consciousness worth its salt is to explain the difference between these major states. As will be seen in Chapter 2, the electromagnetic theory of consciousness eats this requirement for breakfast.
The term "contents of consciousness" refers to the multitude of specific subjective experiences that occur during the states of waking and dreaming. A good theory of consciousness must also make transparent this enormous variety of subjective experiences. This is a much more difficult requirement to fulfill, in part simply because of the vastness of the number of experiences that need to be accommodated. Naturally the electromagnetic theory of consciousness has to work somewhat harder here, but we'll see what it can do.
Simple sensory experiences are probably the easiest kinds of subjective experiences to study, because they have some more-or-less direct relationship with the external world, which allows us to manipulate them experimentally in a more-or-less objective way. Therefore, it may be advantageous to start by searching for common underlying electromagnetic patterns which correlate with sensory experiences. To limn what we are looking for, this may be one or more general electromagnetic patterns that occur across all sensory modalities (with appropriate differences between the senses so that each sensory modality can be understood as generating qualitatively different conscious experiences) plus some more specific spatiotemporal patterns that can be used as building blocks to generate the myriad of different possible specific sensory experiences.
These specific electromagnetic patterns must have a number of characteristics that tie them to sensory experiences:
1.1 - The electromagnetic patterns must be present when and only when particular experiences can be reported by the experimental subject as being conscious. In practical terms, this means that:
1.2 - The electromagnetic patterns must co-vary with subjective experience (rather than with the physical stimulus):
1.3 - The electromagnetic patterns should be generated by localized areas of the brain.
However, the above criterion notwithstanding,
1.4 - The electromagnetic patterns should have a certain global quality, which in practical terms probably means that they should be available over a wide area of the brain.
Above is one set of working criteria by which to judge whether or not certain electromagnetic patterns can be accepted as covarying with conscious experiences. The next step is to review the multitude of experiments already in the scientific literature, to see whether they fulfill the conditions specified for showing correlations between particular states or contents of consciousness and particular electromagnetic patterns generated by the brain. These experiments were probably not originally done with anything like the present hypothesis in mind, but nevertheless they may be the source of a great deal of relevant information. However, it is important to evaluate the experimental methods of these studies quite carefully to ensure that the reported findings are valid.
Broadly speaking there are two methods of measuring spatiotemporal electromagnetic patterns. First, such patterns may be measured by electroencephalographic (EEG) or magnetoencephalographic (MEG) studies, either on human subjects or on animals. EEG and MEG recordings deliver millisecond precision in the time domain, but spatial information is seriously corrupted or smeared by the volume conduction properties of the brain and skull, particularly for EEG measurements. Both techniques measure only contributions from sources that are oriented in a particular way with respect to the surface of the brain. Alternatively, albeit very much less efficiently in terms of patterns generated by the firing of many neurons, spatiotemporal electromagnetic patterns may be measured by multiple single cell recordings from the brains of animals. There are a number practical difficulties with both types of experiments, which must be overcome if the results are to be meaningful. The methodological requirements of neurocognitive studies designed to establish correlations between electromagnetic patterns and cognitive states have been discussed by a number of authors (Donchin, et al., 1977 [76]; Gevins, et al., 1985 [112]; Thatcher & John, 1977 [287]). They include:
(a) - Elimination of artefacts
Raw EEG and MEG recordings are easily contaminated by artefacts arising from muscle activity in the scalp or eye muscles. A particular difficulty is that lateral eye movements (saccades) and blinks have been shown to be correlated with cognitive processes (Just & Carpenter, 1976 [135]; Stern, Walrath, & Goldstein, 1984 [279]). Even instructions to avoid blinking can affect some task-related evoked potentials (Verleger, 1991 [300]). There are various methods in the literature for eliminating such artefacts, some more effective than others (Barlow, 1986 [22]).
Artefacts can also occur during the subsequent analysis of raw EEG data. One commonly used analysis method that is often contaminated by unrecognised methodological artefact is coherence analysis. Coherence measurements reflect the extent to which the oscillations of a particular frequency that are recorded at different electrode sites vary in concert. Being coherent in this sense is thus different from being in phase; e.g. 40 Hz oscillations recorded at two different sites can be out of phase, but still highly coherent if the phase difference between the two stays constant over time. Coherence is currently fashionable because it has recently been proposed as the factor that "binds" together the firing of individual neurons at widely dispersed sites in the brain to produce a single unified conscious percept. The pitfalls in coherence measurement relate largely to EEG recordings, which unlike intracellular recordings are highly dependent on the site of the reference electrode. A basic EEG recording measures the voltage difference between two electrodes, one of which is called the recording electrode and the other the reference electrode. In so-called monopolar recordings, the reference electrode is placed on an ear-lobe (or else a separate electrode is placed on each earlobe and the two are linked electrically) or on the mastoid bone behind the ear and it is therefore assumed that the reference electrode is inactive i.e. not affected by any electric fields produced by the brain. Unfortunately in practice, no reference site is inactive (Lehman, 1987 [163]), (with the possible exception of a non-cephalic or off-head reference, which is then likely to introduce new artefacts arising from the electrical activity of the heart). And depending on the particular reference site chosen, the reference electrode can itself contribute variation to both recordings which significantly influences measured coherence between a pair of electrodes (French & Beaumont, 1984 [99]). There are several explanations for the fact that empirically, coherence measurements do depend on the chosen reference (Beaumont & Rugg, 1979 [24]; Fein, Raz, Brown, & Merrin, 1988 [86]). To cope with this sort of problem, various strategies have been employed to convert EEG measurements to "reference-free" data. Using a computed average of recordings from all the electrodes as a virtual "average reference" is one such strategy, but a better one (because it also sharpens up spatial smearing due to volume conduction through brain and skull tissue) is the use of Laplacian derivatives. There are two main ways of deriving these mathematical transformations and each is subject to controversy (Nunez, et al., 1997 [212]). Life is never simple.
Another possible source of methodological artefact in the recording or analysis of EEG data is the overuse of frequency filters. In general, the wider the bandpass of filters the better, since an overly restrictive bandpass can introduce artefactual waveforms.
(b) - Control of general physiological factors
Brain signals vary with age, gender, handedness, fatigue, habituation, and use of caffeine, nicotine, alcohol and various other drugs. In general, these effects are fairly easy to control, but they must be taken into account. Level of autonomic arousal is another major factor to be considered. General arousal is rather difficult to control and its effects on the electromagnetic patterns generated by the brain are hard to separate out from more specific task-related effects (Kachaturian, Chisholm, & Kerr, 1973 [137]; Kachaturian & Gluck, 1969 [136]). Hence an attempt must be made to control such imprecisely specifiable factors as task difficulty and motivation.
(c) - Isolation of consciousness per se from other specific brain processes
By far the biggest methodological problem with studies designed to reveal the electromagnetic correlates of consciousness is the requirement to isolate the electromagnetic patterns associated with conscious experience from those associated with other brain processes. In general, the requirements for identifying electromagnetic patterns with conscious experiences laid out in the preceding section are aimed at ensuring that the electromagnetic patterns measured correlate with consciousness per se rather than with pre-conscious processing. However there are several issues on which the criteria specified might fall over.
Memory was originally thought to be a single faculty. However most researchers today think of it as a number of systems and subsystems. The classifications used in relation to memory seem to be almost as numerous as the research groups involved, but three stand out.
One of these systems of classification (Kupfermann, 1991 [154]) subdivides memory into declarative memory (which does depend on consciousness, can be laid down after a single event and encodes information about autobiographical events that can be expressed in declarative sentences) and reflexive or procedural memory (which does not depend on conscious processes, builds up slowly over many repetitions and is expressed as improved task performance). Reflexive memory can be demonstrated in animals as uncomplicated as the sea slug. In humans, memories usually begin as declarative memories but may then be transformed by constant repetition into reflexive memories (as for example when a task like driving a car eventually becomes largely automatic and unconscious).
In a second conceptualization, memories are classified according to how long ago they were encoded. They can be iconic (up to about a second after the event) (Breitmeyer & Ganz, 1976 [34]), short-term (a limited capacity type that lasts up to a few minutes) or long-term (apparently unlimited capacity, duration up to a life-time) (Kupfermann, 1991 [154]).
A third classification system (Tulving & Schacter, 1990 [296]) holds that there are three kinds of memory: procedural (which underlies changes in skilled performance), semantic (which has to do with acquisition of factual knowledge in the broadest sense) and episodic (which enables people to remember personally experienced events).
The term working memory fits none of these classification systems exactly. It is term that is often used quite imprecisely, which probably encompasses iconic and short-term memory of the declarative, procedural, semantic or episodic types, together with an attentional component (Baddeley, 1992 [19]).
The mechanism by which memories are laid down is not clear, despite a monumental and exponentially increasing amount of work on the subject over the past few decades. One of the pioneering speculations on the subject was made by the famous neuroanatomist Ramon y Cajal, who suggested that information could be stored by modifying the connections between communicating nerve cells, in order to form associations (Cajal, 1911 [46]). This idea was refined in the late 1940s by Konorski (Konorski, 1948 [148]) and Hebb (Hebb, 1949 [122]). The latter's formalization of the idea, which is known as the Hebb rule and has achieved almost cult status among neuroscientists, is
"When an axon of cell A is near enough to excite cell B and repeatedly and persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased."
The impetus for the current enormous popularity of Hebb's rule was the discovery (Lømo, 1966 [176]) of a real life cellular process in the central nervous system that appears to embody the rule perfectly. This process is now generally called long-term potentiation (LTP) of synaptic transmission. LTP was first described in detail in the hippocampus (Bliss & Gardner-Medwin, 1973 [28]; Bliss & Lømo, 1973 [29]) but it has subsequently been found in every area of the central nervous system in which it has been sought. Its inverse, long-term depression (Pockett & Lippold, 1986 [226]) has also turned out to be widespread and is similarly important for neural network models of memory. A further level of complexity has been introduced by the recent discovery that both processes are themselves plastic - that is, neural activity not only produces immediately observable LTP and/or LTD (depending on the pattern of the activity and the synapses which were active), but also changes the ability of a synapse to undergo LTP and LTD in the future (Abraham & Bear, 1996 [1]).
In sea slugs and snails, which are animals simple enough for the entire neural circuitry that mediates certain behaviors to be known in detail, it is likely that potentiation and depression of transmission at various specific synapses is not only necessary but also sufficient to account for simple forms of learning (Alkon, 1983 [11]; Kandel, et al., 1983 [138]). Basic mechanisms tend to be conserved in evolution, so this, along with a good deal of evidence that manipulating LTP does change learning ability in mammals (Castro, Silbert, McNaughton, & Barnes, 1989 [49]; Morris, Anderson, Lynch, & Baudry, 1986 [201]) suggests that synapses are indeed the cellular location of the changes causing at least some sorts of learning and memory. However the question of where in the brain memories are encoded by such synaptic changes is a vexed one. Here the possibility that different sorts of memories are encoded in different ways becomes relevant.
The original experiments on the subject of memory location were done by Karl Lashley, who ablated various areas of cortex in experimental animals in a search for the "engram", or site of memory storage (Lashley, 1950 [158]). Lashley never found his engram and this failure lead to various models of memory that incorporated an almost mystical property of global storage, such as the holographic ideas of Karl Pribram (Pribram, 1991 [230]). However recent single cell experiments in animals and brain imaging studies in humans show that distinct cortical areas are active during the perceptual and the working memory phases of a memory task (Goldman-Rakic, 1997 [115]). Specifically, perception is correlated with activation of the sensory area of cortex specific to the sense modality used in the task, plus an area of anterior prefrontal cortex. Working memory (which is defined operationally in this case as the memory used to store information in the delay periods of a delayed matching-to-sample task) is correlated with activity only in a particular area of posterior prefrontal cortex. The fact that this area of posterior prefrontal cortex is actually active during the memory phase of the task suggests that working memory may not be due to long-lasting synaptic changes, but simply to continued neural activation. This idea is supported by the fact that gross disruption of neural activity during electroconvulsive therapy destroys memory for events that immediately precede the treatment but leaves older memories intact. The capacity of working memory seems to be limited to approximately seven separate items of information (Miller, 1956 [192]); there is no good neurophysiological explanation for this at present, but the limited size of the cortical area that apparently subserves working memory may be relevant.
Longer term memories probably do involve synaptic modification, in different areas of brain from that involved in working memory. The region that is most implicated in the formation or laying down of long-term memories is the hippocampus. Bilateral loss of hippocampal function in humans causes a severe amnesic syndrome in which working memory is preserved, all varieties of memories that were laid down before hippocampal destruction can still be accessed and new motor skills can still be learned; but no new medium-to-long-term declarative memories can be formed (Milner, Corkin, & Teuber, 1968 [193]; Scoville, 1968 [254]). This clearly indicates that long-term memories are not stored in the hippocampus, but that some form of processing by the hippocampus is necessary for formation of long-term declarative memories. The detailed explanation of this is completely obscure. What is clear though, is that whatever the role of the hippocampus in the encoding or formation of long-term memories, imaging studies show that long-term memory is stored in a distributed cortical system, in which information about specific features is stored close to the regions of cortex that mediate the perception of these features (Ungerleider, 1995 [299]).
The very brief overview above suggests that working memory may well be due to continuing activation of neurons in a specific region of the posterior prefrontal cortex, while long-term declarative memory is encoded by the hippocampus and probably stored all over the brain, in a fashion that closely parallels the distributed representation of perception per se.
In order to subtract the contribution of working memory from the electromagnetic pattern correlating with conscious sensory perception, it may be sufficient to subtract the pattern occurring during the delay phase of a delayed matching-to-sample task from the pattern occurring during the initial perceptual phase. This manoeuvre would at least be a good start. In the absence of some strategy like this one, it is inevitable that the processes underlying working memory would to some degree contaminate attempts to elucidate the electromagnetic pattern corresponding to conscious perception per se.
With regard to long-term memory, similar subtractive paradigms would be necessary (but in this case probably not sufficient) to isolate the electromagnetic patterns correlating with perception per se from those due to involuntary memory formation or recall. For example, comparison of the electromagnetic patterns generated by the brain during direct sensory perception of a particular object and during recall of that object from long-term memory could be a method of removing the effects of recall (although this method would obviously not deal with recall of any other object or thought). Examination of the electromagnetic patterns of a person without a functioning hippocampus would certainly eliminate the effects of long-term memory storage processes.
As with memory, the mechanism of attention is one of the most studied problems in cognitive neuroscience. However it is probably fair to say that our understanding of attention is even less clear than our understanding of memory. At least we can begin with a relatively clear definition: for convenience we will define as "selective attention" any process, voluntary or involuntary, conscious or unconscious, that shapes the selection of one from two or more competing potentially conscious experiences. The broadness of this definition may admit several mechanisms (see below).
Two main metaphors have been proposed in the search for a model of how the brain achieves the outcome of selective attention. The first is a spotlight or searchlight metaphor - the desired information is illuminated by a spotlight and thus made visible or accessible to consciousness (Crick, 1984 [64]; Posner, 1980 [229]; Shulman, Remington, & McLean, 1979 [261]; Tsal, 1983 [294]). The second is a filter metaphor - basically the unwanted information is filtered out (Broadbent, 1958 [37]; Cheal, 1997 [54]; LaBerge, 1995 [156]). Neither metaphor as currently stated directly addresses the question of what decides the aim of the spotlight or the shape of the filter, but since we are for the moment concerned only with simple consciousness, we will continue the robust tradition of glossing over this issue.
Some experimental evidence easily fits into the spotlight metaphor: allocation of attention to a particular location in the visual field improves perception at that location, for example. However because a simple spotlight is sometimes insufficient to explain experimental data, several modifications of this metaphor have evolved. The spotlight has been hypothesized as more like a zoom lens that could vary in diameter (Eriksen & StJames, 1986 [80]), or as forming a gradient away from one or more areas of concentration (LaBerge & Brown, 1989 [155]). Attention gating is another concept along these lines, defining temporal rather than spatial characteristics of the spotlight (Reeves & Sperling, 1986 [235]). A neurophysiological mechanism for implementing the spotlight or searchlight has been proposed by Crick (Crick, 1984 [64]), who suggests that the expression of the searchlight (or multiple searchlights) may be the production of rapid bursts of firing in a subset of thalamic neurons, which in turn act on selected cortical neurons to facilitate formation of transient "cell assemblies". This speculation has not so far been tested experimentally.
Intuitively appealing as the spotlight metaphor may be, however, (particularly in conjunction with the "theatre" metaphor of consciousness in general (Baars, 1997a [17])) it has so far failed to vanquish its major competitor, the filter metaphor. The problem is that the two metaphors are difficult to distinguish experimentally. Both of them in effect predict that wanted information should be represented in the brain strongly and unwanted information represented weakly, or not at all.
The filter metaphor proposes that some sort of matched filter is used for rejecting unwanted information (LaBerge, 1995 [156]) (and in some versions of the scheme, also for facilitating the entry of desired information to consciousness (Cheal, 1997 [54])). Any such filter must be an adaptive one, because there must exist a facility for changing it in accordance with an internal template that specifies what is to be attended to. Thus the filter metaphor of attention also intimately involves memory. Each time a stimulus is presented, it must be measured against some sort of neural template, which has to be retrieved from an updatable context system or memory store.
Brain imaging techniques and data from neurological deficits in attention suggest the existence of four differently located neural networks in the brain subserving the process of selective attention. It is tempting to speculate that these different attentional networks may implement fundamentally different mechanisms, although there is presently little evidence to support this. The effects of these attentional processes are clearly to be found in any brain area that subserves whatever function or percept is the object of selective attention at any given time.
The networks subserving the process of selective attention include:
The cingulate and posterior parietal systems are also sometimes described as involving areas of prefrontal cortex which are concerned with working memory. This suggests either that the task used did not sufficiently distinguish between working memory and attention, or that the filter metaphor (which as already pointed out involves working memory more than does the spotlight metaphor) may be the more relevant to these attentional systems.
It may in fact be the case that the spotlight metaphor applies to the thalamic attention system and the filter metaphor to the cingulate/parietal systems. This idea is given some support by the finding (Portas, et al., 1998 [228]) that attention-related thalamic activity is seen more under conditions of low arousal than under conditions of high arousal. This suggests that the main attentional function of thalamocortical loops may be to selectively modulate the excitability of the cortex. Perhaps when cortical neurons are relatively hyperpolarised, the attentional role of the thalamus may be to "spotlight" various general cortical areas in order to bring them closer to the threshold for firing action potentials (see Appendix B), so that finer-tuned, more delicate facilitation provided by the filter-based systems of the cortical attention areas can be effective.
The role of the cerebellum in all this is presently entirely unclear. One group (Akshoomoff, Courchesne, & Townsend, 1997 [7]) suggests that the cerebellum is "a master computational system that anticipates and adjusts responsiveness to a variety of brain systems (including attention) to efficiently achieve goals determined by cerebral and other subcortical systems". Who knows.
Whatever the mechanism by which they work, it is clear that the final effect of all of these systems for directing attention is to make one particular feature of the environment "take over" consciousness, at the expense of other features. Imaging studies show that the neural events underlying this take place in whatever area of sensory cortex is dedicated to the particular sense involved: i.e. in auditory cortex for auditory tasks (Belin, et al., 1998 [25]; Frith & Friston, 1996 [103]; Fujiwara, Nagamine, Imai, Tanaka, & Shibasaki, 1998 [104]; Pugh, et al., 1996 [231]; Tzourio, et al., 1997 [298]), in the visual cortex for visual tasks (Buchel & Friston, 1997 [42]; Fink, Dolan, Halligan, Marshall, & Frith, 1997 [88]; Haug, Baudewig, & Paulus, 1998 [121]; Hillyard, Vogel, & Luck, 1998 [126]; Watanabe, et al., 1998a [304]; Watanabe, et al., 1998b [305]) and in the somatosensory cortex for somatosensory tasks (Mima, Nagamine, Nakamura, & Shibasaki, 1998 [194]). The modulation caused by attention seems to be to enhance activity due to the attended inputs and this enhancement is uniformly found to occur in both primary and secondary sensory areas; i.e. in the visual system, as early in the processing chain as V1.
It should be clear from the above that it is not possible to separate the effects of attentional processes from conscious perception per se. Conscious perception is only possible when some degree of attention is being paid to the object of perception, and imaging studies show that the effects of the attentional process are inextricably intertwined with the neurophysiology of perception. Whether it is possible in principle to remove the contribution of the processes directing attention to the overall electromagnetic pattern correlating with perception is difficult to say. Apparently the contribution of thalamic mechanisms can be removed if the overall level of arousal is already high. Likewise it has been shown that some tasks do not use the cingulate system. Whether perception is possible without activation of the parietal attentional network is doubtful, however.
In order to establish the fundamental principle about the nature of consciousness that is being proposed here, the first step is to determine whether or not there do exist spatiotemporal electromagnetic patterns that covary with the states and contents of consciousness. This chapter has laid down a fairly demanding set of criteria by which to judge experimental evidence on the issue. In the next four chapters, these criteria are used to examine a number of experiments already described in the scientific literature which may be taken as supporting our fundamental proposal.