In this chapter we apply the contrastive analysis strategy to the neural basis of conscious experience. That is, we look for populations of neurons that control the difference between conscious and unconscious states --- most obviously sleep, waking, and coma. These neural structures behave in several ways like the global workspace model we have developed so far.
There is a curious traditional dichotomy between psychologists and neuroscientists in the way they tend to regard the nervous system. By and large, neuroscientists tend to see a gigantic assemblage of complex neurons, extremely densely interconnected, operating in parallel and at a fairly fast rate (e.g. Thompson, 1967; Shepherd, 1983). Psychologists have traditionally seen a very different system. Their nervous system was slow, appeared to do comparatively simple tasks with high error rates, and seemed to operate serially, performing only one task at a time (e.g. Newell and Simon, 1974; Atkinson & Shiffrin, 19xx; Broadbent, 1958; Norman, 1976). Naturally there are exceptions to these generalizations. Over the past decade‹j ‹ psychologists have increasingly explored parallel or parallel- interactive processes, while some neuroscientists have been studying relatively serial aspects such as event-related potentials (e.g. Anderson, 1983; Hinton and Anderson, 1981; Donchin, McCarthy, Kutas, & Ritter, 1983). Nevertheless, over the broad sweep of the last hundred years of research, the dichotomy between these two views of the nervous system appears to hold to a remarkable degree.
In fact neither extreme is wrong, though both are incomplete. Viewed at the level of neurons, a structure such as the cerebral cortex is indeed immensely complex, containing by recent estimates 55,000,000,000 neurons, each firing off an electrochemical pulse 40 - 1000 times per second, with rich subcortical and contralateral connections, and all apparently active at the same time (Mountcastle, 1978). But when we look at the same system functionally, through input and output performance, it appears to solve simple problems (especially novel ones) at a rate slower than 10 Hz, it makes numerous errors, it tends to serialize even actions that seem superficially executable in parallel, and its efficiency in learning new facts and strategies seems relatively unimpressive.
The difference is, of course, that most psychologists work with the limited capacity component of the nervous system, which is associated with consciousness and voluntary control, while neuroscientists work with the "wetware" of the nervous system, enormous in size and complexity, and unconscious in its detailed functioning. But what is the meaning of this dichotomy? How does a serial, slow, and relatively awkward level of functioning emerge from a system that is enormous in size, relatively fast- acting, efficient, and parallel? That is the key question.
One guise in which this puzzle appears is the issue of " attention" îvsï. "cortical arousal". Both of these concepts have been associated with conscious processes, but in quite different ways (Scheibel, 1980). The psychologist can easily find î selectivityï in human information processing, so that the great array of potential stimulation is reduced to just one stream of information at a time. From William James to the present, psychologists have thought of attention and consciousness in terms of selectivity, a îreductionï in complexity. But the neuroscientist, looking at the nervous system more directly, finds plentiful evidence for system-wide îcortical arousalï associated with wakefulness and orienting to novel stimuli, but much less evidence for selectivity (Shepherd, 1983). Cortical arousal involves widespread desynchronization in the EEG. That is to say: when novel stimuli "catch the attention" of an animal, regular, relatively slow brain waves are interrupted by fast, irregular, low-voltage activity suggestive of increased information processing. This implies not a reduction but an î increaseï in complexity at the neural level. Thus attention and arousal seem to be quite different things, and tend to be treated as separate though somehow related topics. ‹j ‹å This chapter pursues the hypothesis that the split between psychologists and neuroscientists in looking at the nervous system reflects the global-workspace architecture. One advantage of the GW model is that it predicts îbothï selectivity îandï widespread activation, so that it reconciles these apparently contradictory views within a single framework.
The various parts of the nervous system operate all at the same time, and to a degree independently from each other (Thompson, 1976). Further, there is extensive evidence that anatomical structures in the brain often subserve very specialized functions (e.g. Luria, 19xx; Geschwind, 1979). Under these circumstances it is natural to think of the brain as a parallel distributed system, and several interpreters of brain function have done so. Arbib has for some years argued that motor systems should be viewed as collections of multiple specialized processors, operating independently of each other to a considerable degree (e.g. Arbib, 1980). And recently a number of neuroscientists have interpreted the columnar organization of the cerebral cortex in terms of distributed "unit modules" (Mountcastle, 1978; Edelman, 1978). Rozin (1976) has interpreted the evolution of intelligence as an increase in the accessibility of specialized functions, which originally developed as very specific evolutionary adaptations. In more highly evolved nervous systems, he suggests, specialized functions can become available for new adaptive purposes. All these contributors support the idea of the nervous system as a parallel distributed system. Thus Mountcastle (1978) writes:
"The general proposition is that the large entities of the nervous system which we know as the dorsal horn, reticular formation, dorsal thalamus, neocortex, and so forth, are themselves composed of local circuits. These circuits form modules which vary from place to place ... but which are at the first level of analysis similar within any large entity. ... The closely linked subsets of several different large entities thus form precisely connected, distributed systems; these distributed systems are conceived as serving distributed functions." (p. 36).
Mountcastle also interprets the cerebral neocortex as such a collection of specialized distributed processors. The cortex is really a huge layered sheet folded into the upper cranium. Seen in cross-section, this sheet consists of many microscopic columns of cells: ‹j ‹å "The basic unit of operation in the neocortex is a vertically arranged group of cells heavily interconnected in the vertical axis ... and sparsely connected horizontally.
"I define the basic modular unit of the neocortex as a minicolumn. It is a vertically oriented cord of cells ... (which) contains about 110 cells. This figure is almost invariant between different neocortical areas and different species of mammals, except for the striate cortex of primates, where it is 260. Such a cord of cells occupies a gently curving, nearly vertical cylinder of cortical space with a diameter of about 30 microns. ... the neocortex of the human brain ...contains about 600 million minicolumns and on the order of 50 billion neurons.
Next, Mountcastle suggests that these minicolumns of cells are gathered together into îcortical columnsï, which constitute the basic "unit modules" of the cerebral cortex:
"... it is possible to identify within the neocortex a much larger processing unit than the minicolumn. The diameters or widths of this larger unit have been given as 500 microns to 1,000 microns for different areas. ... this larger unit may vary in its cross- sectional form, being round, or oval, or slablike in shape. ... one can estimate that the human neocortex contains about 600,000 of these larger (cortical columns), each packaging several hundred minicolumns. The calculations ... are given to indicate order of magnitude only.
" ... Thus a major problem for understanding the function of the neocortex ... is to unravel the intrinsic structural and functional organization of the neocortical module.
"That module is, I propose, what has come to be called the î cortical columnï."
Unlike Mountcastle, who defines a module anatomically, I would like to view the basic units as functional rather than anatomical (Luria, 19xx). These approaches are not contradictory of course, because functional units must ultimately make use of anatomical units. But there is a difference of emphasis. To mark the difference, I will call these specialized distributed units " processors" rather than "modules".
What part of the brain could carry out the functions described by Model 1? We can specify some of its properties:‹j ‹å First, it should be associated with conscious functions like wakefulness, focal attention, habituation, and indeed all the facts described in the contrastive analyses in this book.
Second, it should fit the model developed in Chapter 2. On the îinput sideï, many systems should have access to the presumed global workspace, and incompatible inputs should compete for access. On the îoutput sideï, it should be able to distribute information to many specialized parts of the nervous system. Since a great many parts of the nervous system seem to be specialized in some way, GW output should be able to reach essentially everywhere.
There is an anatomical and functional system in the brain stem and forebrain that is known to have close relationships with consciousness, in the sense that people gain or lose consciousness when it is activated (Magoun, 1963; Scheibel & Scheibel, 1967; Dixon, 1971; Hobson & Brazier, 1982). This structure includes the classic Reticular Formation discovered by Moruzzi and Magoun (1949), which receives information from all major structures within the brain, including all sensory and motor tracts, and permits very close interaction between all these sources of information. It extends well upward to include the non-specific nuclei of the thalamus. It makes functional sense to include in this larger system the Diffuse Thalamic Projection System, which sends numerous fibers to all parts of the cortex (Figure 3.12). It is possible that cortico- cortical connections should also be included. We will refer to this whole set of anatomical structures as the îExtended Reticular-Thalamic Activating Systemï (ERTAS).
------------------------------ Insert Figure 3.12 about here. ------------------------------
We can summarize the results of a great deal of research since the later 1940s in the following contrastive table:
Stimulation of the Rapid lesioning of the reticular formation reticular formation and the and outer thalamus. outer thalamus, and of the thalamo-cortical projection system.
The lower component of this system, the Reticular Formation of the brainstem and midbrain, was described by one of its co- discoverers as follows:
"Within the brain, a central transactional core has been identified between the strictly sensory or motor systems of classical neurology. This central reticular mechanism has been found îcapable of grading the activity of most other parts of the brainï ... it is proposed to be subdivided into a grosser and more tonically operating component in the lower brain stem, subserving global alterations in excitability, as distinguished from a more cephalic, thalamic component with greater capacities for fractionated, shifting influences upon focal regions of the brain.
"In its ascending and descending relations with the cerebral cortex, the reticular system is îintimately bound up with and contributes to most areas of nervous activity.ï It has to do significantly with the îinitiation and maintenance of wakefulnessï; with the îorienting reflexï and îfocus of attentionï; with îsensory control processesï including habituation ...; with îconditional learningï; through its functional relations with the hippocampus and temporal cortex, with îmemory functionsï; and through its relations with the midline thalamus and pontile tegmentum, with the cortex and îmost of the central integrative processesï of the brain." (Italics added.) (Magoun, 1964)
The fact that the Reticular Formation involves wakefulness, the orienting response, focus of attention, and "most of the central integrative processes of the brain" certainly suggests that it may be a part of what we are looking for. Other‹j ‹ neuroscientists associate parts of this system with the capability of "altering the content of consciousness" (Livingston, 1969), and with "general alerting" and "focused attention" (Lindsley, 1969). The Reticular Formation, which is part of the larger Reticular-Thalamic System we are considering here, thus easily meets our first criterion, that our neuronal candidate should be closely associated with conscious experience.
îNeurophysiological evidence that specialists can cooperate and compete for access to a central integrative "blackboard".ï
The Reticular Formation is called "reticular" (i.e. network- like) because the neuronal axons in this system are usually very short, suggesting a great amount of interaction between adjacent neurons. Further, it receives input from all sensory and motor systems, as well as from other major structures in the brain. Through its connections with the thalamus, it can send information to, and receive it from, all areas of the cortex. If the Extended Reticular-Thalamic System corresponds to our " blackboard", different specialized systems can have access to it.
Aristotle's "common sense" was supposed to be a domain of integration between the different senses. In fact, anatomists who have studied the Reticular Formation have pointed to its resemblance to Aristotle's concept. Scheibel and Scheibel (1967) point out that "Anatomical studies of Kohnstamm and Quensel, which suggested pooling of a number of afferent and efferent systems upon the reticular core, led them to propose this area as a 'centrum receptorium', or 'sensorium commune' --- a common sensory pool for the neuraxis."
Moreover, and of great significance to our discussion, these authors note that "... the reticular core mediates specific delimitation of the focus of consciousness îwith concordant suppression of those sensory inputs that have been temporarily relegated to a sensory role"ï (p. 579). Along similar lines, Gastaut (1969) describes the brain stem reticular formation as an area of "convergence ... where signals are concentrated before being redistributed in a divergent way to the cortex". Thus different sensory contents can suppress each other, as we would indeed expect of input to a global workspace. This meets our second requirement, that different specialized processors can compete for access to the ERTAS.
îNeurophysiological evidence that integrated, coherent information can be broadcast by the Reticular-Thalamic System to all parts of the nervous system.ï
As we noted above, we are including in the term îExtended Reticular-Thalamic Systemï the diffuse thalamic projection system, a bundle of neurons which projects upward like a fountain‹j ‹ From the thalamus to all parts of the cortex. It contains bo specific and non-specific projections, and the specific ones usually contain feedback loops going in the opposite direction as well. The thalamic portion of this system may "broadcast" information from the Reticular System to all parts of the brain. We have already discussed evidence from evoked potentials which indicates that non-habituated stimuli are indeed broadcast non- specifically throughout the brain (Thatcher & John, 1977) (2.xx).
In one possible scenario, one sensory projection area of the cortext provides input input to the Extended Reticular-Thalamic Activating System. If this input prevails over competing inputs, it becomes a global message which is widely distributed to other areas of the brain, including the rest of the cortex. Thus one selected input to the ERTAS is amplified and broadcast at the expense of others.
We can therefore suggest that the ERTAS underlies the " global broadcasting" function of consciousness, while a selected perceptual "processor" in the cortex supplies the particular î contentsï of consciousness which are to be broadcast. (These are typically perceptual contents, because the ERTAS receives collateral pathways from all sensory tracts; and of course, we have previously remarked on the favored relationship between conscious experience and perception). These conscious contents, in turn, when they are broadcast, can trigger motoric, memory, and associative activities.
There is independent evidence that cortical activity îby itselfï does not become conscious (Magoun, 1964; Libet, 1977, 1978, 1981; Shevrin and Dickman, 1980). We would suggest that any cortical activity must trigger ERTAS "support" in a circulating flow of information, before it can be broadcast globally and become conscious (e.g. Shevrin and Dickman, 1980; Scheibel, 19xx). Dixon (1971) has also argued that a circulating flow of information between the reticular formation and the sensory areas of the cortex is required before sensory input becomes conscious.
There are probably several ways to gain access to the brain equivalent of a global workspace. In one scenario, two perceptual inputs arrive in the cortex at the same time through the direct sensory pathways, and begin to compete for access to the limited- capacity system --- presumably the thalamus and reticular formation. Suppose the two events are auditory and visual, so that we get stimulus competition (Figure 3.13). One may be a speech sound in the left ear, and the other a falling glass in the right visual field. It has been known for at least a century‹j ‹ that two simultaneous, incompatible events do not become conscious at the same time (e.g. Wundt, 1912; Blumenthal, 1977). In our scenario, only one of the two can be broadcast at any moment, because they conflict in spatial location and content, so that the two simultaneous cortical events cannot be fused into a single, consistent conscious event. One of the two may be favored because of readiness in the receiving specialized processors to support it. For instance, we may be alert to the possibility of the glass falling; in that case, the specialized processors involved with moving a hand to catch the falling glass would trigger quickly to help direct consciousness to the visual stimulus, and away from the auditory input. Possibly there is rapid alternation between the visual and auditory stimulus, so that each is broadcast for 100 milliseconds to recruit additional processors. Receiving processors may then support the visual message over the auditory one. But glasses fall quickly; losing a few hundred milliseconds will probably cause us to miss the falling glass; and competition for access to consciousness inevitably slows down effective action.
This scenario has the following features. First, there is competition between perceptual systems for access to the global workspace. Only one input can win, and it is the one that garners most support from potential receiving systems, especially those that are ready for and "interested in" the winning system. " Winning" means that one system gains access to the thalamus and perhaps reticular formation, allowing a general broadcasting of at least some central aspects of the winning system --- perhaps its spatiotemporal properties, its significance, its relevance to current goals, etc. Probably some receiving processors gain more information from the global message than others. There is probably a circulating flow of information between the winning input system, the global workspace, and the receiving processors, each component feeding back to the others, so that there is for some time a self-sustaining loop of activated systems (see Figure 3.21). Possibly this flow may allow more direct local channels to be established between the perceptual input and some receiving systems; over time, this local flow of information may allow the creation of a new, efficient specialized "falling glass detector," which operates independently of the global workspace.
------------------------------------ Insert FIG. 3.13 about here ------------------------------------
‹j While the neurophysiology seems compatible with the GW model, it also suggests some additions to the model.
1. îThe outer thalamus as a common sensory mode.ï
The outer layer of the thalamus, the nucleus reticularis thalami, is thought to contain a body-centered spatio-temporal code, that can "gate" different inputs before cortical activation occurs (Scheibel, 1980). Thus auditory signals to the right rear of the body may be coded in one place, and visual signals in the same location may converge on the same area. This suggests the existence of a kind of lingua franca in which the outer thalamus may thus act as a common sensory mode. The thalamic centers have much more specificity in this sense than the lower reticular centers. m
2. îThe brainstem reticular formation as a mode switch. ï
What then, is the role of the Reticular Formation (RF) --- especially the brain stem components that are known to be involved in sleep, waking, and coma? The RF may act as a "mode switch" on the system that does more specific selection. If we use the search-light metaphor of consciousness, the RF nuclei may act as a dimmer switch, to increase or decrease the amount of light, but not to direct it to any particular object. In terms of the GW model, the RF may control overall activation levels, while the thalamic nuclei may modulate activation to and from specific specialized processors.
3. îLocations of some specialized capacities.ï
îSensory/imaginal systems as GW input.ï A large part of the cortex is devoted to perceptual analysis, especially vision, and this may be one reason for the predominance of perceptual/imaginal input to consciousness. It seems likely that imagery also makes use of these perceptual systems, with stimulation of internal origin. Thus some of the input specialists would seem to be located in the sensory projection areas of the cortext.
Clearly voluntary decisions can affect conscious contents, and these are not perceptual for most people, so that it is possible that non-perceptual events can gain global access. Alternatively, it is possible that these non-perceptual systems‹j ‹ make use of perceptual/imaginal processors to gain access to the system underlying consciousness.
îShort term memory and the hippocampus.ï
There is now good evidence that the hippocampus, a structure that surrounds the thalamus, is closely associated with the transfer of short-term memory information to long term memory (e.g. Milner, 19xx). Clearly short-term memory is intimately associated with consciousness, and if the hippocampus contains such a system, it is presumably one of the recipients of global broadcasting (Winson, 19xx).
îVoluntary speech control and the rehearsal component of short term memory.ï
Similarly, voluntary control of speech is clearly involved in short-term rehearsal, as in memorizing a telephone number. Speech production is one of the few functions that is quite well lateralized to the left hemisphere (Springer & Deutsch, 19xx), in particular to Broca's area. It seems likely that this system is involved in mental rehearsal, which is after all mental speaking; rehearsal really acts to refresh conscious access to immediate memory. Therefore this rehearsal system would also seem to provide input to the GW. However, voluntary control in general is more associated with the frontal cortex, so that this functional system may include both frontal areas and Broca's area.
4. îSpatio-temporal coding as a lingua franca.ï
We have claimed that perception and consciousness have a special relationship, in the sense that all qualitative experiences are perceptual or quasi-perceptual (like imagery or inner speech). All perceptual experiences involve spatio-temporal information, of course, and the neurophysiology indicates that a great many neural systems can process spatio-temporal information. This suggests that spatio-temporal coding may be one lingua franca that is broadcast through the neural equivalent of a global workspace.
5. îGlobally broadcast information may feed back to its sources.ï
‹j ‹å If broadcasting is truly global, the systems that provide global input should also receive their own results, just as a television playwright may watch his own play on television. Such a circulating flow back to the source is postulated in certain cognitive theories. It is known to have a number of useful properties. For example, McClelland & Rumelhart (1981) have shown that a circulating flow in an activation model of word recognition helps to stabilize the representation of the word.
6. îReceivers of global information may feed back their interest to the global workspace.ï
The physiological evidence discussed above suggests that global îoutputï flows in two directions as well. There are anatomical connections that allow feedback from cortex back to the thalamus. Such feedback loops are extremely common in the nervous system. Most sensory systems allow for a flow of information "top-down" as well as "bottom up." This anatomical evidence may mean that receiving systems, those that take in globally broadcast information, may be able to feed back their interest to the global workspace, thus strengthening or weakening any particular global message. One can make an analogy to the well-known Nielsen Ratings for television programs in the United States. Each program is continuously sampled to see how many viewers are watching it, and programs of low popularity are quickly dropped. In a later chapter we will suggest that this kind of popularity feedback may explain such phenomena as habituation and the development of automaticity with practice (Chapter 5).
7. îOther anatomical systems may facilitate global broadcasting.ï
The diffuse thalamic projection system (Figure 3.12) is not the only projection system that may be used to broadcast information. There are long tertiary cortical neurons that connect frontal to other areas of the cortex, and cross- hemispheric fibers that connect the two halves of the cortex through the corpus callosum. All such transmission pathways may be involved in global broadcasting.
8. îCyclical snowballing rather than immediate broadcasting.ï
The neurophysiology suggests that broadcasting may not be an‹j ‹ instantaneous event, but a "snowballing" recruitment of global activation, supported by many systems, that may feed back on each other. For example, Libet's work indicates that for cortical activity to become conscious may take as long as a half second (Libet, 1978; 1981). This is much longer than a single broadcast message would take, and suggests a circulating flow between cortical and sub-cortical areas, building upon itself until it reaches a threshold. Thus we must not take the broadcasting metaphor too literally: a relatively slow accumulation would accomplish much the same functional end. This kind of snowballing would of course also explain the role of the anatomical feedback loops described above.
9. îAttention: Control of access to the global activating system.ï
Later in this book we will draw a distinction between consciousness and îattentionï --- in which the latter serves to control access to consciousness. Such attentional systems have been found in the parietal and frontal cortext (e.g. Posner, 1982). Possibly the frontal components are involved in voluntary control of attention, which can often override automatic attentional mechanisms (see Chapter 8).
Figure 3.21 is a modified version of Model 1, with feedback loops from the global message to its input sources, and from the receiving processors back to the global message. We will find additional evidence for these feedback loops later in this book.
v * v ---------------------©©©©©©©©©É Insert Figure 3.21 about here. ---------------------©©©©©©©©©
The above interpretation of the neurophysiology resembles to‹j ‹ earlier models of the Reticular Formation (RF), which we treat here as a subset of the more broadly defined ERTAS system (Moruzzi & Magoun, 1949; Lindsley, 19xx; Magoun, 1964). Arguments for a central role in conscious experience of the RF have come under some criticism (e.g. Thompson, 1967; Brodal, 1981). Some of these criticisms serve to qualify our conclusions, though they do not contradict them decisively.
îFirstï, as more detailed studies have been performed using long©term implanted electrodes, a number of specific components have been found in the RF, so that the bald statement that the RF is nonspecific is not quite true (Hobson & Brazier, 19xx). We should be careful not to refer to the whole RF and thalamus as subserving these functions, but only to nuclei and networks within these larger anatomical structures. îSecondï, under some circumstances one can show that lesioned animals with little or no surviving RF tissue show relatively normal waking, sleeping, orienting and conditioning. It is possible that the outer layer of the thalamus may be able to replace RF functions, especially if the lesions are made gradually, so that there is time for adaptation to take place. îThirdï, it is clear that a number of other parts of the brain are involved in functions closely related to conscious experience, such as voluntary attention; the sense of self; voluntary control of inner speech, imagery, and skeletal musculature; and control of sleep and waking. We must be careful therefore not to limit our consideration to just the extended reticular-thalamic system; surely many other systems act to contribute to, control, and interact with any neural equivalent of a global workspace.
Before concluding this chapter, we should mention the puzzling role of brain duality. The human brain has a major division down the midline, extending far below the great cortical hemispheres into most subcortical structures, including the thalamus and even the brainstem reticular formation. This suggests that duality may be an "architectural" feature of the nervous system. But Model 1 has no place for duality; it emphasizes unity rather than duality.
Brain duality is a fundamental fact of nervous system functioning. In the intact brain, it is not clear that it has major functional implications; most of the evidence for brain lateralization in normal people shows only very short time delays between left and right-sided functioning. The corpus callosum, which connects the two hemispheres, is estimated to add perhaps 3 milliseconds of transmission time to interactions between the two sides --- not enough to make much of a difference (D. Galin, personal communication, 1986). Still, this massive anatomical feature must be functional in some sense, and it is curious that our architectural approach to the nervous system has no obvious role for it. It is possible that its role is primarily‹j ‹ developmental, and that in the intact adult brain its effects are more difficult to observe (e.g. Galin, 1977).
Even with these qualifications, the evidence is strong that parts of the Extended Reticular-Thalamic System serve as major facilitators of conscious experience, while the cortex and perhaps other parts of the brain provide the îcontentï of conscious experience. This evidence can be naturally interpreted in terms of the GW model, derived from purely cognitive evidence. Contributions from both the ERTAS and cortex are presumably required to create a stable conscious content. The evidence comes From numerous studies showing a direct relationship between t ERTAS and known conscious functions like sleep and waking, alertness, the Orienting Response, focal attention, sharpening of perceptual discriminations, habituation of orienting, conditioning, and perceptual learning (see references cited above). Further, there is evidence consistent with the three major properties of Model 1: first, that major brain structures, especially the cortex, can be viewed as collections of distributed specialized modules; second, that some of these modules can cooperate and compete for access to the ERTAS; and third, that information which gains access may be broadcast globally to other parts of the nervous system, especially the huge cortical mantle of the brain.
Thus substantial neurophysiological evidence seems to be consistent with Model 1, with one addition: there is evidence of a feedback flow from cortical modules îtoï the ERTAS, suggesting that a circulating flow of information may be necessary to keep some content in consciousness. In addition, global information may well feed back to its own input sources. Both kinds of feedback may serve to strengthen and stabilize a coalition of systems that work to keep a certain content on the global workspace. These modifications have has been incorporated into Model 1A (Figure 3.21).
Let us review where we have been. First, many neuroscientists suggest that the nervous system is a distributed parallel system, with many different specialized processors. A constrastive analysis of neurophysiological evidence about conscious vs. unconscious phenomena focused on the well-known reticular formation of the brainstem and midbrain, on the outer layer of the thalamus, and on the diffusely projecting fibers From the thalamus to the cortex. Several established facts abo the nervous system suggest that we may take the notion of îglobalï‹j ‹ broadcasting quite seriously, that conscious information is indeed very widely distributed in the central nervous system. At least parts of the reticular-thalamic system bear out our expectations regarding a system that can take input from specialized modules in the brain and broadcast this information globally to the nervous system as a whole.
‹ ‹ Footnotes
Footnote 1. I am most grateful to several neuroscientists and psychologists who have provided valuable comments on the ideas presented in this chapter, especially David Galin, Benjamin Libet, and Charles Yingling (UCSF), Paul Rozin (U. Pennsylvania), Michael A. Wapner (CSULA), Theodore Melnechuk (Western Behaviorial Sciences Inst.), and Arnold Scheibel (UCLA).