The Wright Institute

Berkeley, Calif. 94704

bbaars@wrightinst.edu

In Max Velmans (Ed.) (1996) The Science of Consciousness: Psychological, Neuropsychological, and Clinical Reviews. London, Routledge.

1.0 Introduction.
2.0 A little history.
3.0 What evidence is relevant to consciousness AS SUCH?
4.0 Operational definitions:
Conscious, unconscious, and fringe conscious.
5.0 Treating consciousness as a variable.
6.0 Pros and cons of conscious and unconscious processes.
7.0 Consciousness has limited capacity but provides
GLOBAL ACCESS to system capabilities.
8.0 Some metaphors for consciousness.
9.0 Modern theories.
10.0 The many functions of consciousness.
11.0 Summary and conclusions.

1.0 Introduction.

No alien space visitor could fail to observe that vertebrates,
including humans, engage in purposeful motion only two-thirds of the
earthly day. In the remaining third, we hibernate. When consciousness
returns in the morning, a massive change in electrical activity takes place
all over the cortex, as the fast, small, and irregular waves of waking EEG
replace the large, slow, regular hills and valleys of deep sleep.
Accompanying this faster electrical activity, we humans report a rich and
varied array of conscious experiences: colors and sounds, feelings and
smells, images and dreams, the rich pageant of everyday reality. Because
these reports of conscious experience covary so perfectly with the pattern of electrical activity, we routinely infer that they reflect a single underlying reality. In this broad sense, the centrality of consciousness is pretty much beyond reasonable doubt.

At this instant you, the reader, are conscious of some aspects of the
act of reading --- the color and texture of THIS PAGE, and perhaps the
inner sound of THESE WORDS. But you are probably not aware of the
touch of your chair at this instant; nor of a certain background taste in your mouth, nor that monotonous background noise, the soft sound of music, or the complex syntactic processes needed to understand THIS PHRASE; nor are you now aware of your feelings about a friend, the fleeting events of several seconds ago, or the multiple meanings of ambiguous words, as in THIS CASE. Even though you are not currently conscious of them, there is good of evidence that such unconscious events are actively processed in your brain, every moment you are awake.

When we try to understand conscious experience we aim to explain
the differences between these two conditions: between the events in your
nervous system that you can report, act upon, distinguish, and
acknowledge as your own, and a great multitude of sophisticated and
intelligent processes which are unconscious, and do not allow these
operations.

2.0 A little history.

Our concern with consciousness is not new. At the very dawn of
written thought Plato compared human experience to the shadows
projected on the wall of a cave by people walking and gesturing in front of
a fire. The mind, he thought, was like a bound prisoner, thrown on the
cave floor with his face to the wall, able to see the fire-cast shadows, but ignorant of the underlying reality. While Plato's Allegory of the Cave is ancient, its implications are quite modern --- from the vast unknown of the cave, to the striking capacity limits of conscious experience, and the
helplessness of the prisoner. Very similar ideas are found in the vast Asian libraries on human consciousness, and they are common in contemporary thinking as well.

3.0 What evidence is relevant to consciousness AS SUCH?

To study anything in science we need to treat it as a variable. The
concept of gravity would have been useless had Newton been unable to
imagine zero gravity. Likewise, to understand consciousness we need to
compare its presence and absence; consciousness without unconsciousness
is meaningless. Only if we can compare conscious and unconscious events
can we ask, WHAT IS THE DIFFERENCE BETWEEN THE TWO? Only
then can we deal with the issue of consciousness AS SUCH.

George Mandler has made the penetrating observation that science
is obliged to treat consciousness not as an observable datum but as an
inferred concept based on public evidence. As scientists dealing with public
evidence and shared reasoning, we observe only the REPORTS people
make about their conscious experience. In a vast number of cases we can
make quite reliable inferences about human experience based on such
public reports. We now have almost 200 years of evidence that under well-defined conditions perceptual reports correspond with exquisite sensitivity to the physical stimulus array. In studying perception, in practice we always treat verbal reports as representations of private experience. Entire domains of research, like perception and psychophysics, depend upon this methodology.

Of course, what we infer from experiential reports is not just
anything: the experiences people report are overwhelmingly ABOUT
something. They are representations. Conscious experience is, after all,
the condition under which people can describe and respond
discriminatively to all the different aspects of an experienced event.

UNCONSCIOUS representations can also be inferred from public
observations, though people cannot act upon them deliberately. The
simplest example is the great multitude of memories that are currently
unconscious. You may recall this morning's breakfast --- but what
happened to that memory before it was brought to mind? We believe it was
still represented in the nervous system, though not consciously. For
example, unconscious memories can influence other processes without
ever coming to mind. If you had orange juice for breakfast today, you may
try milk tomorrow, even without bringing today's juice to mind. A case can
be made for unconscious representation of habituated stimuli, memories
before and after recall, automatic skills, implicit learning, the rules of
syntax, unattended speech, presupposed knowledge, pre-conscious input
processing, and many other phenomena. Researchers still argue about the
particulars of some of these cases, but it is widely agreed that given
adequate evidence, unconscious representations may be inferred.

It is crucial to treat both conscious and unconscious representations
as inferred constructs, for only then can we treat consciousness as a
variable, allowing us to talk about consciousness AS SUCH. As in the case
of Newtonian gravity we can compare a state with its absence. This is a bit more abstract than the familiar experimental method, because are
comparing two INFERRED entities, rather than two direct observations.
But the principle of treating consciousness as a variable to be studied is
exactly the same.

Notice how often in the history of science, the ability to treat
something as a variable created a breakthrough. When finally Newton
was able to imagine a space without friction, contrary to 20 centuries of
Aristotelian physics, he could write his laws of motion. An object in motion continues in motion. But on earth there is no such thing as frictionless space. It required a great act of imagination for natural philosophers were able to treat friction as a variable with a zero point. The same great leap of imagination made all the difference in determining the nature of the atmosphere, of gravity on earth, of oxygen as the crucial component of the air we breathe, of absolute zero degrees temperature, and so on. Contrastive analysis aims to do just that with respect to consciousness. It is not a new idea in science. All the more reason to try it.

Using contrasive analysis we can define a large set of empirical
constraints on the concept of consciousness. We can compare not just
single phenomena like implicit and explicit memory, but entire sets of
contrastive pairs of phenomena in various domains (see Tables 1 and 2).
Together the contrasts constrain any theory of consciousness in an
empirically solid and demanding way. Very few theories can satisfy all of
these constraints. (See Baars, 1988.)

4.0 Operational definitions: Conscious, unconscious, and fringe conscious.

Conscious processes can be operationally defined as the set of eventsthat:

(a) are claimed by people to be conscious;
and which
(b) can be reported and acted upon,
(c) with verifiable accuracy, (1)
(d) under optimal reporting conditions.

"Optimal reporting conditions" implies a minimum delay between
the event and the report, freedom from distraction, and the like. All this fits standard practice in the study of perception, short-term memory, problem- solving, imagery, and many other phenomena.

There is a curious asymmetry between the assessment of conscious
and unconscious processes. Obtaining verifiable experiential reports
works very nicely for specifying conscious events; but unconscious ones are much more slippery. In many cases of apparently unconscious processes, such as all the things the reader is not paying attention to at this moment, it could be that the "unconscious" events may be momentarily conscious, but so quickly or vaguely that we cannot recall it even a fraction of a second later.

Or suppose people cannot report a word shown for a few milliseconds: does this mean that they are truly unconscious of it? William James understood this problem very well, and suggested in fact that there were no unconscious psychological processes at all (1890/1983, p. 162ff.). We have called this the "zero point" problem (Baars, 1988; see Holender, 1986).

This is one of those tricky cases in which the criterion for unconsciousness could retreat ever further and further beyond the grasp of diligent experimenters. Jacoby and Kelley (1992) suggest an attractive
answer, a criterion for unconscious events which does not solve the
problem exactly, but which gives a reasonable basis for consensus.
Suppose, they suggest, we ask a subject to consciously AVOID reporting
recently learned memory when a cue that tends to trigger the material is
presented? If they cannot do so and the memory comes to mind unbidden,
the processes involved are plausibly unconscious.

To illustrate the point we can ask the reader to fill in the missing
words or syllables as quickly as possible, but not to mention any male
individual in that process: mother/ ....... ; sister/ ...... ; aunt/ ....... ; cow/
......... ; stallion/ ........... ; Norse/........... ; George Washing/ ............... ;
Saturday / ......day ; Abraham Lin/.......... ; Thomas Jeffer/ ...........; Lyndon
John/ ......... ; automo/.......... ; Ingmar Berg/...........; garbage/ .......... . If,
in spite of your intention not to fill in completion for males, you said
"NorseMEN," "JefferSON," "JohnSON, BergMAN" etc. the association
to males was plausibly unconscious. While Jaccoby &;Kelley used specially memorized words rather than existing word knowledge, the logic is quite similar.

If William James were to come along at this point to argue that the
fact that the word represented a male could still have been conscious very
fleetingly, we might answer yes, possibly, but in such a way that the
unconscious representation was apparently unable to affect deliberate,
voluntary operations upon the event. This may not be the ultimate
solution; the Jacoby &;Kelley criterion only guarantees that the
"unconscious" event cannot be used in what might be called "practical
consciousness" --- the ability to act upon and report a fleeting mental
event. But distinguishing "practical conciousness" from "practical
unconsciousness" may be the best we can do for the time being.

In sum, mental events can be defined as UNCONSCIOUS for
practical purposes if:

(a) Their presence can be verified (by influencing other
observable tasks, for example); although
(b) they are not claimed to be conscious;
(c) and they cannot be voluntarily reported, acted on, or
avoided;
(d) even under optimal reporting conditions.

By this definition, practiced skills like typing, reading, balancing
one's body, and visual analysis are largely unconscious in their details.
Likewise, blindsight, implicit memory, unattended information, subliminal
effects, the details of language processing, the effects of priming, etc.,
would be unconscious. There is again a reasonable fit between this
definition and existing scientific practice.

Fringe Conscious Events.

There is an interesting class of phenomena that are not quite
conscious nor unconscious, but which are nevertheless very important for
our normal mental functioning. William James thought that "fringe
conscious" events were at least as important as focally conscious
experiences. Indeed, he thought that perhaps one-third of our conscious
lives may be spent in subjectively active but vague states of mind. Fringe
events include feelings of rightness, beauty, coherence, anomaly,
familiarity, attraction, repulsion, and the like.

Most people are sure of their judgment when they experience something as beautiful. But is the experience of beauty specifiable in detail, like the sight of a red toothbrush? Surely not for most people, even when they are very sure about the experience. The combination of high certainty, high accuracy, and low experienced detail defines a "fringe conscious" state.

Mangan (1993) has developed James' ideas about fringe consciousness in modern terms, suggesting that fringe phenomena may not be subject to the classical capacity limitations of conscious experiences. The claim is that feelings of familiarity or coherence can be simultaneously present in consciousness along with perceptual contents, for example. As
we listen to a wonderful melody, we can simultaneously feel moved by it
emotionally, feel that it is familiar, and have a sense of rightness and fit.
Since focal conscious capacity is notoriously limited to one internally
consistent at a time, Mangan sees fringe experience as a means of evading
that limitation. The fringe may be, in Mangan's terms, a "radical
condensation" of unconscious information in consciousness. Fringe states
seem to be very useful. There is evidence that they are involved in accurate
decision-making, predict resolution of tip-of-the-tongue states, and give a sense of availability of a memory even before it comes to mind.

Bowers, Rehger, Balthazard and Parker (1992) asked subjects to
choose a word triplet that has a common associate, a fourth word that is
closely related to the first three words. Quite often subjects will feel
certain that a certain triplet "feels right" without knowing why. Consider
A and B below:

A B
playing still
credit pages
report music

Triplet A has a common associate ("card") while B does not. Subjects can indicate which triad has a solution EVEN WHEN THEY CANNOT PROVIDE THE MISSING WORD. They can "respond discriminatively to coherence that they could not identify". Bowers et al suggest that "this tacit perception of coherence guided subjects gradually to an explicit representation ... in the form of a hunch or hypothesis." (p. 72) Fringe experiences are evidently useful.

Research on fringe consciousness is still in its early stages. We can
however suggest a useful operational definition for fringe conscious
events, as those experiences that:

(a) can be reported by all normal subjects in similar tasks,
(b) with verifiable accuracy and high confidence;
and
(c) which can be voluntarily acted upon,
(d) BUT WHICH ARE NOT CLAIMED TO HAVE DIFFERENTIATED PERCEPTUAL, IMAGINAL OR SEMANTIC CONTENT; even
(f) under optimal reporting conditions.

5.0 Treating consciousness as a variable.

As elsewhere in science, we can only study a phenomenon if we can
treat it as a variable. In the case of consciousness we can compare waking
to sleep, coma, and general anesthesia; subliminal to supraliminal input,
habituated vs. novel stimuli, attended vs. nonattended streams of
information, recalled vs. nonrecalled memories, and the like. Table 1
shows a contrastive analysis for Stimulus Representation, comparing
conscious stimuli to unconscious input processing of the same information--- as in pre-perceptual processes, unattended input, habituated representations, and the like. In all these cases there is evidence that the unconscious stimulus is accurately represented in the nervous system, so that we can compare conscious and unconscious representations of a given input.

Contrastive analysis, as we have said, is much like the experimental
method: We can examine closely comparable cases that differ only in
respect to consciousness, so that consciousness becomes, in effect, a
variable. However, instead of dealing with only one experimental data
set, we can examine entire categories of well-established phenomena,
summarizing numerous experimental studies. In this way we can highlight
the variables that constrain consciousness over a very wide range of cases.

Even a broad Contrastive Analysis brings out some important facts.
Table 1 shows that conscious perception involves more than just an
accurate representation of the stimulus, because there are many kinds of
stimulus representation that are not conscious. Likewise in the case of
imagery, we see that there are conscious analogues of images such as
habituated mental images, which have faded from consciousness but
nevertheless serve to guide task performance.

What is the critical difference then between the two columns of Table 1? There are several underlying differences, but perhaps the most
significant one is that conscious percepts and images can trigger ACCESS
to unanticipated knowledge sources. It is as if the conscious event is
broadcast to memory, skill control, decision-making functions, anomaly
detectors, and the like, allowing us to match the input with related
memories, use it as a cue for a skilled actions or decisions, and detect
problems in the input. At a broad, architectural level, conscious
representations seem to provide access to multiple knowledge source in
the nervous system, while unconscious ones seem to be relatively isolated.
The same conclusion follows from other contrastive analyses. (See Baars,
1988).

6.0 Advantages and disadvantages of conscious and unconscious
processes.

Whatever we do really well, we tend to do unconsciously, from
speaking to seeing to playing piano. This observation has led some
psychologists to wonder why consciousness is needed at all. To get at this
question of the role of consciousness we can conduct another contrastive
analysis, focused on the capabilities of comparable conscious and
unconscious processes. Table 2 presents a basic set of such Capability
Contrasts. Notice that purely conscious processes are handicapped by low
computational efficiency: they are rather inefficient (slow, prone to error,
and vulnerable to interference), serial, and limited in capacity. Consider
mentally multiplying of 23 x 79. For most of us this is not very easy to do,
and the more steps are conscious, the more difficult it is. Yet mental
multiplication is trivial in complexity compared to the vast amount of
processing that is needed to analyze the syntax of THIS SENTENCE. But
syntactic analysis is of course entirely unconscious. Mental arithmetic can become more efficient with practice, by letting highly predictable steps become automatic and unconscious, but that illustrates the same point, that efficiency in computational processes is achieved when some algorithm becomes unconscious.

Purely conscious mental manipulations have a high rate of errors,
are slow, and interfere with each other, suggesting that performing
efficient symbolic computation is not the primary function of
consciousness.

But the computational drawbacks of experience are balanced by clear advantages: consciousness has a VAST RANGE of possible contents, it
enables novel ACCESS to to an astonishing number of skills and
knowledge sources, and it shows EXQUISITE CONTEXT-SENSITIVITY.

As an example of its vast range, consider all the possible percepts,
images, memories, concepts, intentions, fringe experiences, and the like, of which we can be conscious. As one poetic student put it, we can be
conscious of everything from the "rumbling of our stomach to the return of
a theme in a Bach cantata."

A conscious event, like this sentence, can access new information in
memory, combine knowledge from different sources in the brain in novel
ways, and trigger unconscious rule systems that will pick up errors in any
level of analysis --- in the meaning, syntax, word level, sound, intonation,
or printing of this snetnecne. (copy editor: do not change)

The context-sensitivity of consciousness can easily be shown to
permeate whole domains, like perception, thinking, or language. Take the
predominance of lexical ambiguity, the fact that most words in natural
languages have multiple meanings. The Oxford English Dictionary, for
example, devotes 75,000 words to the many different meanings of the little word "set." "Set" can be a verb, noun, or adjective. It can be a game in tennis or a collection of silverware. We can set a value on an antique, or
look at a stage set. Glamorous people make a jet-set, and unlucky ones a
set of fools. The sun sets, but it is certainly not set in concrete.
Mathematicians use set theory, but psychologists talk about set as
readiness to do or experience something.

In the sentences above, notice how quickly we can change our set
about the word "set." In each use of the word you had litttle difficulty
understanding the meaning intended, providing you had sufficient
CONTEXTUAL information to make it fit. But notice that the detailed
process of relating contextual information to the choice of meaning is
rarely conscious. Thus whatever we experience is shaped by unconscious
processes, just like a theater in which we see only the actors on stage, but not all the people behind the scenes who make it all work. If the same
actors were off stage (unconscious) their actions would not be
contextualized by the entire supportive apparatus of the theater. The
context-sensitivity of conscious events extends far beyond language to
perception, action control, memory, problem-solving, etc. But there is no
evidence that novel combinations of subliminal words can be understood
unconsciously. Likewise, when we make a navigational error simply
because we are so used to turning right on the street going home, the less
conscious we are, the more we are likely to make the error. The less
conscious some event is, the less it is sensitive to context.

Unconscious processes have their own advantages, however.
Unconscious automatisms, such as the ones that control nearly all aspects
of the act of reading, show of impressive speed and accuracy in routine
matters, a tendency to perform parallel or concurrent processing
whenever possible, and, when all unconscious resources are taken
together, a vast capacity. But of course there is constant interaction
between conscious and unconscious processes. In listening to a friend
describe last night's party, we follow the conscious flow of sound, words,
and meaning with no awareness of the complex acoustic, phonological,
morphological, lexical, syntactic, semantic, intonational, and pragmatic
process are happening at the same time, all of which are needed for us to
become aware of the message.

Yet we can easily prove that these sophisticated unconscious
processes are going on all the time. Should our friend commit any error,
such as saying "entymology" instead of "etymology" or lapse into German
word order, we would immediately detect the error (if we were paying
attention), even though it occurred at any of a dozen different levels of
analysis. Further, concurrent with all this fast, complex and unconscious
linguistic activity, we also maintain balance and upright posture, represent predictable aspects of all incoming stimuli, and shape our actions in terms of the social and pragmatic demands of the situation.

In the laboratory, the limitations of purely unconscious language
processing have been highlighted in selective attention studies. If we
receive two dense flows of information, like two simultaneous stories, one in each ear, or two different ball games shown on the same television set, we can follow only a single, consistent flow of the action. Under these conditions we can present information to the "unattended channel," the ear one is not listening to, for example. In general, it has been found that semantic priming from individual words in the unattented channel can
influence the experience of of the conscious, attended channel. Thus, the
word "money" in the unattended message can bias understanding of the
word "bank" toward "financial institution" instead of "shoreline of a
river". However, the information in the unconscious channel does not
extend to the meaning of longer passages (see Greenwald, 1992 for a
review).

Consistent patterns of evidence now begin to emerge from the
contrastive analysis. We see the interplay of a serial, integrated, and very
limited stream of consciousness with an unconscious system which is
distributed, composed of autonomous modules, and of enormous collective
capacity.

7.0 Consciousness has limited capacity but provides GLOBAL ACCESS to
numerous mental resources.

Limited capacity.

Many psychologists and philosophers have noted the limitations on
moment to moment conscious capacity. In each conscious moment, we
tend to be conscious of only a single thing, i.e. a scene, intention, or
daydream. In everyday life, we know that we cannot do two things at one
time, such as have an intense conversation and drive in downtown traffic.
If one task does not require much conscious involvement we can carry on
two activities at a time, probably by switching rapidly between the two
tasks.

The number and duration of items CURRENTLY REHEARSED in
working memory is similarly limited; classically we talk about 7 +/- 2
words or numbers in short-term memory, but that number drops to 3 or 4
when we are unable to rehearse them. Likewise intentional (controlled)
actions can only be carried out serially, one at a time, while automatic
processes can be carried out concurrently (LaBerge, 1980). In carefully
studied dual-task situations, consciously controlled tasks interfere with
each other, causing errors and delay. But when one or both tasks become
automated through practice, the interference declines and can disappear
entirely.

In spite of these limitations, consciousness seems to be the gateway
through which we gain access to a vast continent of knowledge, skills, and
actions. The weight of evidence suggests that all learning requires
consciousness of the material to be learned. Even in the case of implicit
learning, where the regularities in the learned material seem to be
unconscious or at least relatively unconscious, we still need to be conscious of the target stimuli that display the unconscious regularities.

Transfer of learning to a new situation requires us to make explicit, conscious connections between the old and the new cases. The example of errors in a written sentence shows again that we need to be conscious of the sentence in order to activate numerous sophisticated unconscious knowledge sources that can signal (consciously) when an error occurs. Further, all perceptual information, the basic source of information about the world, is conscious. And all the "ego functions," such as the ability to make a decision, to act voluntarily, to recall events in memory, and the like, depend upon conscious information.

The global reach of consciousness is shown most easily by biofeedback. There is firm evidence that any single neuron, or any population of neurons, can be controlled using biofeedback signals ---but they must be conscious signals. Subliminal stimulation, being distracted From the signal, being habituated to it --- all these cases prevent of the thumb can tap into a single motor unit --- a muscle fiber controlled by one neuron coming from the spinal cord, and another going back to it. When the signal from the muscle fiber is amplified and played back as a click through a loudspeaker, the subject can learn to control the click within 10 minutes, and some subjects have been able to play drumrolls on their single motor unit after about 30 minutes of practice. That is what we mean when we say that consciousness creates global access to all parts of the nervous system.

The paradox is therefore that conscious contents seem limited, but
somehow give access to the vast array of knowledge sources in the brain.
How can we explain this?

8.0 Some metaphors for consciousness.

Over the last two centuries, scientific thought about consciousness
has often been cast in terms of a small set of metaphors.

This metaphor embodies the commonly held assumption that
consciousness results from the activation of some stimulus or memory
above a certain threshhold value. The hypothesis was stated in an early
form by J. Herbart in 1824. Activation theories of memory and perception
are of course the bread and butter of modern cognitive psychology. They
can explain priming effects, availability of memory, and other forms of
readiness to respond.

A few modern theories have begun to identify activation with access
to consciousness. However a simple activation metaphor fails as an
adequate account of consciousness, because it cannot handle the
phenomenon of redundancy effects: Stimuli or behaviors that are repeated
and predictable become unconscious. This includes habituation of
perceptual events and mental images, skills practiced to the point of
automaticity, and semantic satiation for repeated meaningful words.
Redundancy effects occur in all sensory modalities, probably at all levels of analysis, and apparently even for abstract concepts. If some minimum
amount of activation were all that is needed to enter consciousness, we
would left with the paradox that repeating activation to the point of
redundancy always causes conscious contents to become unconscious!
Simple, unadorned activation cannot explain both the likelihood of a
stimulus coming to mind and also fading from mind if the stimulus
happens to become predictable.

Activation alone is inadequate as a theory of consciousness. We need something more.

The Tip-of-the Iceberg Metaphor

Another long tradition looks at consciousness as the tip of a large
psychological iceberg. This metaphor brings out the fact that conscious
experience emerges from a great mass of unconscious events. Sigmund
Freud is of course the name that comes to mind most easily here, but the
current view is not necessarily psychodynamic. Unconscious material is not unconscious because it has been pushed out of consciousness. Rather, the great bulk of neural systems in cortex and elsewhere are very efficient specialized processors which are normally unconscious. Modern
neuropsychology confirms the vastness of the iceberg that is under the
water, compared to the small limited-capacity component that is visible
and conscious.

The Novelty Metaphor

One line of thought about consciousness holds that consciousness is
focused on novelty, "antihabit," or that which mismatches our
expectations or mental set (Mandler, 1984). There is ample evidence that
people and animals seek novel and informative stimulation; consciousness
seems to have a preference for "news." Repetitive, predictable, "old"
stimuli tend to fade from consciousness regardless of their sensory
modality, degree of abstractness, or physical intensity. Novelty can be
defined as change in the physical environment (dishabituation),
disconfirmation of expectation (surprise), or violation of skilled routines
(choice-points in the otherwise routine flow of action). The Novelty
Metaphor captures one central function of consciousness--- its ability to
direct resources toward adaptation to novel and significant events. In the
language of Piaget, consciousness comes into play more when we need to
accomodate to unexpected events, than when we readily assimilate to
predictable ones.

However, the novelty hypothesis is obviously not enough. We can be conscious of routine matters if they are important enough personally or
biologically: our repetitive need to eat, for example, without getting
habituated of bored.

The Integration Metaphor.

In the early information processing models of cognitive psychology
there was sometimes an implicit assumption that consciousness was the
result of "high level processing." Once low level sensory information was
synthesized through succeeding stages of constructive information
processing, the output could be made available to an "executive" which
handled conscious percepts. Neisser (1967) described multiple, primitive or low-level "pre-attentive processes" operating in parallel which
constructed the potential contents of consciousness. The output of only
one pre-attentive process entered consciousness and the others died out.
For Marcel (1983) and Mandler (1984) conscious perception results from
high level perceptual integration. In addition to theories of perception,
psycholinguistic theories describe complex, hierarchical use of the rules of grammar during language comprehension, with the assumption that only
the output at the highest level enters consciousness.

Several psychologists suggest that conscious contents involve the
highest level of integration in some hierarchical view of the nervous
system. Sensory systems have increasingly integrated and abstract
anatomical maps of the sensory receptor surface. Points of light and light- dark contrast at the retina are integrated at the thalamus, whence it is in turn mapped onto higher levels in the primary visual cortex, representing lines, colors, orientations, and simple shapes, proceeding to at least another 25 increasingly complex and integrated visual maps, until finally we obtain such high-level constructs as three-dimensional object constancy, moving bodies, faces, kinetic (force) representations of objects, and entire, integrated multi-sensory scenes. Consciousness can surely access such high-level information.

In language understanding, we know that adults most of the time
attend to the MEANING of speech, rather than lower levels of analysis
such as pure sound, phonemes, morphemes, words, grammar, and the like.
For instance, try to remember the sentence before this one. In all likelihood you won't remember the sound, syntax, or even the specific words; only the gist, the meaning. This fact is consistent with the idea that you devote conscious capacity to the meaning level of language.

Nevertheless, you CAN choose pay attention to the individual words or speech sounds. When people are presented with paired comparisons between two speech sounds that differ in very tiny acoustical properties, they can generally learn to tell the difference. In fact, in learning a new language and culture we have to pay attention to all levels, and a singer or trained speaker can learn to attend to extremely subtle aspects of speech sounds. Thus consciousness is not limited to the high levels of analysis: it easily ranges across levels.

In perception the same pattern obtains. While we are normally conscious of entire visual scenes, we can be conscious of only a single star on a dark night, even though the starlight may be as small as a single stream of photons streaming into a single retinal receptor, out of all the millions of receptor cells. Consciousness is marked by its FLEXIBILITY in ranging up and down perceptual and language hierarchies. If we encounter a problem in communicating our meaning to a listener, we can quickly restate the same meaning in a very different form. But we can equally well change the fast and complex movements of the tongue to avoid touching a painful tooth. After a several minutes of biofeedback training even the operation of single motor neurons can come under conscious control (without much training people have learned to play drum rolls on a single motor unit). The critical condition for biofeedback is immediate, conscious feedback from the to-be-controlled neurons. Again, it appears that we can get conscious access to quite high levels of perceptual and conceptual analysis, but consciousness does not seem to be limited to high level of analysis.

At the brain level, integration issues are often viewed in terms of the "binding problem." (Crick &;Koch, 1990) That is, how is activity in neurons at diverse cortical locations "bound together" to produce coherent,
integrated, and rapidly changing conscious experiences? Objects and
events have many features that change rapidly, triggering quite different
activity in many different brain locations, sensory modalities, and with
some asynchrony. How are multiple aspects of a stimulus, such as THIS
WORD or SENTENCE brought together into a single, momentary,
coherent conscious experience? Crick and Koch (1990) and others have
suggested that binding may be paced by a neural timing signal, specifically the 40Hz wavefor that can be observed in many places in the brain at the same time. However, recent studies of anaesthetized subjects shows that the 40 Hz waveform can exist without consciousness. The status of the binding problem is an ongoing topic of debate.

The Executive.

Another popular view is that consciousness is closely associated with
executive control over action, speech, and thought. Sometimes this is
identified with a "self as agent,"operating in a hierarchy of control and
monitoring functions (Hilgard, 1992; Baars, 1988). Kihlstrom (1993)
maintains that an activated mental representation of the self, residing in
working memory, is a necessary condition for conscious experience. In this view, for example, autobiographical memory must be indexed as self-
related as it is encoded in the brain. Such self-indexing presumably takes
place in immediate memory. Conversely, material that is not tagged as
self-related cannot be recalled as one's own past experience, though it
may well emerge in tests of implicit memory, learned skills, and the like. In common sense terms, practicing "pinky-trills" on the piano will improve one's playing of Mozart without having to bring up the autobiographical memory of the last time we practiced trills. Indeed, recalling the last time we practiced anything (such as reading sentences like this) would interfere disastrously with our ability to do the action that was practiced.

These ideas are significant because they attempt to bring the self into
our cognitive architecture, which, in its own way is as important as trying to make sense of consciousness. However, there are some unanswered questions: For instance, if conscious experience requires a self-system (at least for registration of autobiographical information) how could we have"self-alien" experiences, like minor dissocations in the case of accidents, when we might experience something in a removed, dreamy, or self-dissociated fashion, as if the accident is happening to someone else; or the case of alien hand syndrome in parietal neglect (where one's hand seems to belong to someone else); or in the kind of self-alien experiences that are routinely used in hypnotic induction? For example, hypnotic techniques typically ask people to experience their hands as rising by themselves, as if drawn by a string. About 20% of the population is highly hypnotizable, and can easily be induced to have a strong self-alien experience of this kind. But surely all these examples are conscious. If these conscious but self-alien experiences exist, conscious contents must not require self-tags to be conscious. This might imply, that such hypnotic experience are less likely to be recalled as autobiographical memories, a perfectly testable proposition.

Shallice (1978, 1991) has presented a very different kind of executive
association with consciousness, one that is exclusively involved with the
control of action. In early versions, the executive supervisor of this system was the conscious component, but more recently Shallice modified the theory to explain consciousness as coherent functioning of a set of
controlling modules. (See below).

The Searchlight Metaphor

One of the ancient, embedded idioms in many languages is consciousness as something that casts a light on things, so as to clarify and enlighten one's understanding. (Ah, I see! ). The image of casting light is discussed as far back as Plato. The Searchlight metaphor is our modern way of expressing these ideas (e.g. Lindsay and Norman (1977) and Crick
(1984)). It is an attractive metaphor, embodying in a single image the
selective function of consciousness, and the flow of conscious contents
across numerous domains of memory, perception, imagery, thought, and
action. Crick's neurobiological version also captures what is known about
the thalamocortical complex, the fact that the thalamus, nestled like an
egg in the cocoon of each hemisphere, maps point-to-point into the
corresponding parts of the cerebral cortex, almost like a miniature brain
within the brain. The searchlight of attention can be imagined as shining
out from the thalamus to the corresponding regions of the cortex.

It must be relevant that the thalamus contains two nuclei (the reticular and intralaminal) whose lesioning uniquely abolishes consciousness. Cortical lesions, on the other hand, only abolish the contents of consciousness, not consciousness itself. Crick's thalamocortical spotlight is an elegant image, and one that may yet turn out to be true.

Both psychological and neurobiological searchlight theories are open
to two unanswered questions: 1) how is the PARTICULAR focus selected?
That is, what determines whether the searchlight shines on THIS content
or cortical area and not some other? And, 2) Once something is selected to
be in focus, what happens to that information? What does it mean for that
content to BE conscious? Is it conveyed to a self-system, as suggested by
the Executive metaphor? Does it go to motor systems, to prepare them for
voluntary action? Or does it go to semantic processors, which encode the
meaning of the event? One can imagine a real searchlight operating in full
darkness, so that one can see the light and its target, but not the people
who control its aim, nor the audience that is looking it. What is happening
in these dark spaces? Without filling them in, the metaphor is missing
some essentials. The Theatre metaphor makes a stab at an answer.

The Theater in the Society of Mind

The theater metaphor likens conscious experience to the brightly lit
stage in a darkened auditorium. Whatever is on stage is disseminated to a
large audience, as well as to the director, playwright, costume designers
and stagehands behind the scenes. The focus of this metaphor is a publicity
function in a vast array of specialized systems, which constitute the
audience. Events on stage are privileged; they are made available to all the
listeners in the auditorium.

Dennett (1991) and Dennett and Kinsbourne (1992) have criticized a
particular version of the metaphor which they call the Cartesian Theater,
"a single place in the brain where 'it all comes together,'" much like the
tiny, centrally-located pineal gland in Descartes' brain. This is a bit of a
red herring, however. No current model suggests that conscious contents
may be found in a single, tiny point, the discrete finish line of competing
potential contents. There are many other ways to bring multiple sources of
information together, for example by coordinating the dozens of
perceptual maps in the nervous system. The thalamus may be in an ideal
situation to do that.

Modern theater metaphors bypass these Cartesian paradoxes. In
recent versions, consciousness is not identified with any single locus.
Rather the contents of consciousness are whatever is being widely
disseminated throughout the brain. (Newman &;Baars, 1994).

Combining the metaphors into a single, coherent theory.

We can combine all five metaphors into a single integrated
"supermetaphor." The theater can visualized to include useful aspects of
the threshold, searchlight, iceberg, novelty, and the executive notions. As
such a supermetaphor becomes enriched, it can gradually take on the
features of a genuine theory. The theoretical proposals described below
can be seen as steps in that direction.

9.0 Modern Theories of Consciousness

We are now beginning to see a small cluster of first-approximation
theories that aim to account for a number of aspects of conscious
experience.

9.1. Johnson-Laird: a computational view.

Johnson-Laird's (1988) operating system model of consciousness
emphasizes the control aspects, such as directing attention, planning and
triggering action and thought, and purposeful self-reflection. Johnson-
Laird proposes that the cognitive architecture s parallel processing system dominated by a control hierarchy. The system is a collection of largely independent processors (finite state automata), which cannot modify each other but which can receive messages from each other; each starts to compute when it receives appropriate input from any source. Each passes messages up through a hierarchy to the operating system, which sets goals for the subsystems. The operating system does not have access to the detailed operations of the subsystems--- it receives only their output. Likewise, the operating system does not need to specify the details of the goals it transmits to the processes--- these take the goal, abstractly specified, and elaborate it in terms of their own capabilities.

In this model, conscious contents reside in the operating system or itsworking memory. Johnson-Laird believes his model can account for
aspects of self-reflection, intentional decision-making, and action control.

9.2. Schacter's Model of Dissociable Interactions and Conscious
Experience (DICE).

Accumulating evidence regarding neuropsychological disconnections of processing from consciousness, particularly implicit memory and anosagnosia, led Schacter to propose his Dissociable Interactions and Conscious Experience (DICE) model. "The basic idea motivating the DICE model ... is that the processes that mediate conscious identification and recognition -- that is, phenomenal awareness in different domains --- should be sharply distinguished from modular systems that operate on linguistic, perceptual, and other kinds of information" (p. 160-161, 1990).

Like Johnson-Laird, Schacter's DICE model assumes independent
memory modules and a lack of conscious access to details of
skilled/procedural knowledge. It is primarily designed to account for
memory dissociations in normally functioning and damaged brains. There
are two main observations of interest. First, with the exception of coma
and stupor patients, failures of awareness in neuropsychological case are
usually restricted to the domain of their impairment; they do not have
difficulty generally in gaining conscious access to other knowledge sources. Amnesic patients do not necessarily have trouble reading words, while alexic individuals don't necessarily have memory problems.

However, implicit (nonconscious) memory for unavailable
knowledge has been demonstrated in many conditions. For example,
name recognition is facilitated in prosopagnosic patients when the name is accompanied by a matching face--- even though the patient does not
consciously recognize the face. Numerous examples of implicit knowledge
in neuropsychological patients who do not have deliberate, conscious
access to the information are known (see Milner and Rugg, 1992). These
findings suggest an architecture in which various sources of knowledge
function somewhat separately, since they can be selectively lost; these
knowledge sources are not accessible to consciousness, even though they
continue to shape voluntary action.

In offering DICE, Schacter has given additional support to the idea
of a conscious capacity in a system of separable knowledge sources,
specifically to explain spared implicit knowledge in patients with brain
damage. DICE does not aim to explain the limited capacity of
consciousness or the problem of selecting among potential inputs. In
agreement with Shallice (see below) the DICE model suggests that the
primary role of consciousness is to mediate voluntary action under the
control of an executive. However, the details of these abilities are not
spelled out, and other plausible functions are not addressed.

9.3. Shallice's Supervisory System.

Shallice shares an interest in the relationship of volition and
consciousness with James (1890), Baars (1988), and Mandler (1984). His
earlier theory (1978) focused on conscious selection of the DOMINANT
ACTION SYSTEM, the set of current goals that work together to control
thought and action. More recently Shallice (1988; Norman and Shallice,
1980) has modified and refined the theory to accomodate a broader range
of conscious functions. Shallice describes an information-processing system with five characteristics.

First, it consists of a very large set of specialized processors like
Johnson-Laird's subsystems and Baars' (1988) specialized unconscious
processors. There are several qualifications on this "modularity":

a. there is considerable variety in the way the subsystems can
interact;

b. the overall functional architecture is seen as partly innate partly
acquired, as with the ability to read;

c. the "modules" in the system include not only input processors but
also specialized information stores, information management specialists,
and other processing modules.

Second, a large set of action and thought schemata can "run" on the
modules. These schemata are conceptualized as well-learned, highly
specific programs for routine activities, such as eating with a spoon,
driving to work, etc. Competition and interference between currently
activated schemata is resolved by another specialist system, CONTENTION SCHEDULING, which selects among the schemata based on activation and lateral inhibition. Contention scheduling acts during routine operations.

Third, a SUPERVISORY SYSTEM functions to modulate the
operation of contention scheduling. It has access to representations of
operations, of the individual's goals, and of the environment. It comes
into play when operation of routinely selected schemata does not meet the
system's goals, that is, when a novel or unpredicted suituation is
encountered or when an error has occurred.

Fourth, a LANGUAGE SYSTEM is involved which can function
either to activate schemata or to represent the operations of the
supervisory system or specialist systems.

Fifth, more recently an EPISODIC MEMORY component
containing event-specific traces has been added to the set of control
processes.

Thus, the supervisory system, contention scheduling, the language
system, and episodic memory all serve higher level or control functions in
the system. As a first approximation, one of these controllers or several
together might be taken as the "conscious part" of the system. However,
as Shallice points out, consciousness cannot reside in any of these control
systems taken individually. No single system is either necessary or
sufficient to account for conscious events. Consciousness remains even
when one of these control systems is damaged or disabled. And, the
individual control systems can all operate autonomously and
unconsciously. Instead, Shallice suggests, consciousness may arise on
those occasions where there is concurrent and coherent operation of
several control systems on representations of a single activity. IN THIS
EVENT, THE CONTENTS OF CONSCIOUSNESS WOULD CORRESPOND TO THE FLOW OF INFORMATION between the control systems and the flow of information and control from the control systems to the rest of the cogntive system.

Shallice's (1988) model aims primarily to "reflect the phenomenological distinctions between willed and ideomotor action (p. 319)." However, the move to identify consciousness with the control of coherent action subsystems and the emphasis on the flow of information among the subsystems, bears a strong resemblance to the information
broadcasting ideas of Baars' global workspace theory.

9.4 Baars' Global Workspace Theory.

The theatre metaphor is the best way to approach Baars' Global Workspace (GW) theory (Baars, 1983, 1988). Consciousness is associated with a global "broadcasting system" that disseminates information widely throughout the brain. If this is true, then conscious capacity limits may be the price paid for the ability to make single momentary messages available to the entire system for purposes of coordination and control. Since at any moment there is only one "whole system," a global dissemination facility must be limited to one momentary content. (There is evidence that each conscious "moment" may be on the order of 100 milliseconds, a tenth of a second).

Baars develops these ideas through seven increasingly detailed
models of a Global Workspace architecture, in which many parallel
unconscious experts interact via a serial, conscious, and internally
consistent Global Workspace (or its functional equivalent) (1983, 1988).
Global workspace architectures have been developed by cognitive
scientists since the 1970's, and the framework is closely related to the well-known integrative theories of Herbert A. Simon, Allan Newell, and John R. Anderson. Architectures much like this have also seen some practical applications. GW theory is currently the most thoroughly developed framework, aiming to explain by far the largest set of evidence. It appears to have fruitful implications for a number of related topics such as spontaneous problem-solving, voluntary control, and even the Jamesian"self" as agent and observer.

GW theory relies on three theoretical constructs: expert processors, a Global Workspace, and contexts. The first construct is the specialized
unconscious processor, the "expert". We know of hundreds of types of
"experts" in the brain. They may be single cells, such as cortical feature
detectors for color, line orientation, or faces, but also entire networks and systems of neurons, such as cortical columns, functional areas like Broca's or Wernicke's, large nuclei such as locus ceruleus, etc. Like human experts, unconscious expert processors may sometimes be quite narrowminded. They are extremely efficient in limited task domains, able to act independently or in coalition with each other. Working as a coalition, the do not have the narrow capacity limitations of consciousness. They can receive global messages, and if they can mobilize a coalition of other experts, they may be able to control a perceptual processor that can place a mental image, phrase of inner speech, or even perceptual content in consciousness. For routine missions they may work autonomously, without conscious involvement, or they may display their output in the global workspace. Answering a question like "What is your mother's maiden name?" requires a misson-specific coalition of unconscious experts, which report their answer to consciousness.

The second construct is of course the Global Workspace (GW) itself. A GW is an architectural capability for system-wide integration and
dissemination of information. A Global Workspace is much like the podium at a scientific meeting. Groups of experts may interact locally around conference tables, but in order to effect change any expert must compete with others, perhaps supported by a coalition of experts, to reach the podium, whence global messages can be broadcast. New links between
experts are made possible by global interaction via the podium, and can
then spin off to become new local processors. The podium allows novel
expert coalitions to form, to work on new or difficult problems, which
cannot be solved by established experts and committees. Tentative
solutions to problems can then be globally disseminated, scrutinized, and
modified.

It is clear that the evidence presented in Tables 1 and 2 above falls
into place by assuming that information in the Global Workspace
corresponds to conscious contents. Since conscious experience seems to
have a great perceptual bias, it is convenient to imagine that perceptual
processors --- visual, auditory, or multimodal --- can compete for access
to a brain version of a GW, but of course perceptual input systems may in
turn be controlled by coalitions of other experts.

Obviously the abstract GW architecture can be realized in a number
of different ways in the brain, and we do not know at this point which
brain structures provide the best candidates. While its brain correlates are
unclear at this time, there are possible neural analogues, including the
reticular and intralaminar nuclei of the thalamus, one or more layers of
cortex, or an active loop between sensory projection areas of cortex and
the corresponding thalamic relay nuclei. Like other aspects of GW theory,
such neural candidates provide testable hypotheses. (Newman &;Baars,
1993)

"CONTEXT," the third construct in GW Theory refers to the powers
behind the scenes of the theater of mind. Contexts are coalitions of expert
processors that provide the director, playwright, and stagehands behind
the scenes of the theater of mind. They. can be defined functionally as
STRUCTURES THAT CONSTRAIN CONSCIOUS CONTENTS WITHOUT BEING CONSCIOUS THEMSELVES, just as the playwright determines the words and actions of the actors on stage without being visible. Conceptually, contexts are defined as pre-established expert coalitions that can evoke, shape and guide global messages without themselves entering the global workspace.

Contexts may be momentary, as in the way the meaning of the first
word in a sentence shapes an interpretation of a later word like "set," or
they may be long-lasting, as with life-long expectations about love,
beauty, relationship, fate, pride, and all the other things people care
about. While contextual influences shape conscious experience without
being conscious, contexts can be SET UP by conscious events. The word
"tennis" before "set" shapes the interpretation of "set," even when "tennis"
is already gone from consciousness. But "tennis" was initially conscious,
and needed to be conscious in order to create the unconscious context that
made sense of the word "set."

Thus conscious events can set up unconscious contexts. The reader's ideas about consciousness from years ago may influence his or her current experience of this chapter, even if the memories of the earlier thoughts do not become conscious again. Earlier experiences typically influence current experiences as contexts, rather than being brought to mind. It is believed for example that a shocking or traumatic event earlier in life can set up largely unconscious expectations that may shape subsequent experiences. (Baars & McGovern, in press).

The three constructs of Global Workspace Theory --- expert processors, Global Workspace and contexts --- are specifically defined with respect to conscious and unconscious functioning, unlike other psychological concepts, but of course they have many resemblances to well-known ideas. Figure 3/4 shows a summary of these similarities.

The example of goal contexts.

The reader is doing a number of purposeful things right now, but they are not necessarily conscious, not at least unless they are interrupted. Suppose, for example, a flyspeck drifted into your visual field right now, making the next word impossible to read. Obviously you would do something to remove it, because it was interfering with reading, even
though you probably did not hold the goal of reading in consciousness just
a minute ago. Nevertheless, something goal-like was happening, why else
would you remove an obstacle to reader? Moreover, your goal of reading
this paragraph is probably nested under higher-level goals, which you
could bring to mind but which were not necessarily conscious either, until
you were reminded of them. Further, there are likely to be high-level goals--- such as pursuing a decision made years ago to explore interesting
psychological topics like consciousness --- which are helping to guide your actions without being conscious. And finally, every word above can be
viewed as a subgoal serving a higher-level goal, so that we are talking
about a dense hiarchy of goals and subgoals, many of which were
conscious, or consciously accessible at some point, some of which may be
brought to mind when they are disrupted, and some of which may be
entirely unconscious. This is essentially the rationale for the concept of a
goal hierarchy (not in principle different from Maslow's motivational
hierarchy); Global Workspace theory simply suggests that those goals are
contextual when they shape action and experience. That is, the goal
hierarchy helps to direct and shape a complex act like reading without
itself being conscious. Figure 5 represents a simplified goal context
hierarchy for the sentence "I want to buy ice cream," a goal that is of
course nested within longer lasting higher levels goals, such as
maximizing pleasure and avoiding pain.

Formally we can define goal contexts as future-directed, non-qualitative representations about one's own actions (Figure 4). Like other contexts, they shape and constrain conscious contents, including possible conscious goal-images, without being conscious themselves. In everyday usage, a goal context is the same as a plan or intention. Goal contexts are similar to the notion of "current concerns" (Singer, 1988) which is a conscious or unconscious need or value that shapes attentional selection, imagery, inner speech, and action, including daydreams, sampled stream of thought, and even night dreams. By definition they exert their influence on conscious experience unconsciously, unless they themselves become the
objects of experience (when they are of course shaped by yet other
contexts). We tend to become aware of the workings of goal contexts only
when they are unexpectedly mismatched, for example, when we make
errors, or otherwise cannot complete our intentions.

The GW notion of goal context hierarchy differs from the goal hierarchies of Maslow and others in only one major respect: These theories have little to say about whether goals can be conscious.Note that in GW Theory, consciousness is found in the Global Workspace and not in the high levels of the goal hierarchy. From this perspective, the remarkable thing about consciousness is not that it is always "high level," but rather that it can range freely among levels of goals and in the goal context hierarchy.

10.0 Functions of consciousness.

William James wrote that "The PARTICULARS OF THE DISTRIBUTION OF CONSCIOUSNESS, so far as we know them, POINT TO ITS BEING EFFICACIOUS ...." (1890/1983, Vol. I, p. 141-2). This view contrasts markedly with epiphenomenalism, which, in the words of T. H. Huxley, views conscious experience much as the "steam whistle which accompanies the work of a locomotive [but which] is without influence upon its machinery". (quoted in James, 1890/1983, Vol. I, p. 130).

The evidence sketched above militates strongly against epiphenomenalism. If consciousness is a major biological adaptation, it may have not just one but a number of functions. The bloodstream circulates oxygen and glucose to all the body cells, takes away waste products, provides a channel for hormones, carries the white cells of the immune system, plays a role in temperature regulation, and much more. Fundamental biological adaptations tend to accrue multiple functions.

The evidence suggests at least the following functions for conscious
experience:

By relating global input to its contextual conditions, the system
underlying consciousness acts to define a stimulus and remove ambiguities in its perception and understanding.

The more novelty the nervous system must adapt to, the more
conscious involvement is required for successful learning and problem-
solving.

3. Prioritizing and Access control function.

Attentional mechanisms exercise selective control over what will
become conscious. By consciously relating some event to higher-level
goals, we can raise its access priority, making it conscious more often and therefore increasing the chances of successful adaptation to it. By
persuading smokers that the apparently innocuous act of lighting a
cigarette is life-threatening over the long term, the medical profession has raised the smokers' conscious involvement with smoking, and created the possibility for more creative problem-solving in that respect.

Conscious goals can recruit subgoals and motor systems in order to
organize and carry out voluntary actions.

5. Decision-making and Executive function.

While the global workspace is not an executive system, executive
access to the GW creates the opportunity for controlling any part of the
nervous system, as shown by the extraordinary range of neural
populations that can be controlled with conscious biofeedback. When
automatic systems cannot resolve some choice-point in the flow of action, making it conscious helps to recruit knowledge sources able to help make the proper decision. In the case of indecision, we can make a goal conscious to allow widespread recruitment of conscious and unconscious resources acting for or against the goal.

Conscious goals and plans are monitored by unconscious rule
systems, which will act to interrupt execution if errors are detected.
Though we often become aware of making an error in a general way, the
detailed description of what makes an error an error is almost always
unconscious.

Through conscious inner speech and imagery we can reflect upon
and to some extent control our conscious and unconscious functioning.

Automatized, "canned" responses are highly adaptive in predictable
situations. However, in facing unpredictable conditions, the capacity of
consciousness to recruit and reconfigure specialized knowledge sources is
indispensible.

In sum, consciousness appears to be the major way in which the
nervous system adapts to novel, challenging, and informative events in the
world.

11.0 Summary and conclusions.

A vast array of solid evidence is beginning to reveal the role of
consciousness in the nervous system, at least in broad outline. Conscious
experience seems to create access multiple, independent knowledge
sources. While organization of coherent percepts and control over novel,
voluntary actions may have been primary in the phylogenetic evolution of
consciousness, it seems also to have acquired other functions which can be seen as contributing to adaptive action in a complex world, such as self-monitoring and self-reflection, symbolic representation of experience, control over novel actions, and mental rehearsal.

Reference Notes

(1) The verifiability criterion may exclude phenomena like hallucinations and dreams, which are hard to verify in detail. While these are important, we exclude them initially in order to focus on cases that command the widest agreement.

References

Baars, B. J. (1983). Conscious contents provide the nervous system with
coherent, global information. In R. Davidson, G. Schwartz, &;D. Shapiro
(Eds.), CONSCIOUSNESS AND SELF-REGULATION (pp. 45-76). New York: Plenum Press.

Baars, B. J. (1988). A COGNITIVE THEORY OF CONSCIOUSNESS. New York: Cambridge University Press.

Baars, B.J. &;McGovern, K.(in press) CONSCIOUSNESS & COGNITION.

Bowers, Rehger, Balthazard and Parker (1992) Intuition. COGNITIVE PSYCHOLOGY.

Crick, F. (1984). Function of the thalamic reticular complex: The
searchlight hypothesis. PROC. NAT. ACAD. SCI. USA, 81, 4586-4590.

Crick, F., and Koch, C. (1990). Towards a neurobiological theory of
consciousness. SEMINARS IN NEUROSCIENCE, 2, 263-275.

Dennett, D. (1991). CONSCIOUSNESS EXPLAINED. Boston: Little, Brown.

Dennett, D., &;Kinsbourne, M. (1992). Time and the observer: The where and when of consciousness in the brain. BRAIN AND BEHAVIORAL SCIENCES, 15(2).

Flanagan, O. (1993) CONSCIOUSNESS RECONSIDERED. Cambridge, MA: MIT Press.

Greenwald, A. (1992). New Look 3, Unconscious cognition reclaimed. AMERICAN PSYCHOLOGIST, 47(6), 766-779.

Hilgard, E. R. (1992). Divided consciousness and dissociation.
CONSCIOUSNESSAND COGNITION, 1(1), 16-31.

Holender, D. (1986). Semantic activation without conscious identification in dichotic listening, parafoveal vision, and visual masking: A survey and appraisal. BEHAVIORAL AND BRAIN SCIENCES, 9, 1-66.

Jacoby, L. L., &;Kelley, C. M. (1992b). A process-dissociation framework for investigating unconscious influences: Freudian slips, projective tests, subliminal perception, and signal detection theory. CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE, 1(6), 174-179.

James, W. (1890/1983). THE PRINCIPLES OF PSYCHOLOGY.
Cambridge, MA: Harvard University Press.

Johnson-Laird, P. N. (1988). A computational analysis of consciousness. In A. Marcel &;E. Bisiach (Eds.), CONSCIOUSNESS AND CONTEMPORARY SCIENCE (pp. 357-368). Oxford: Oxford University Press.

Kihlstrom, J. F. (1993). The psychological unconscious and the self. In G. R. Bock &;J. Marsh (Ed.), CIBA SYMPOSIUM ON EXPERIMENTAL ANDTHEORETICAL STUDIES OF CONSCIOUSNESS, (pp. 147-156). London: Wiley Interscience.

Kinsbourne, M. (1993). Integrated field model of conscousness. In G. R. a. M. J. Brock (Eds.), CIBA SYMPOSIUM ON EXPERIMENTAL AND THEORETICAL STUDIES OF CONSCIOUSNESS (pp. 51-60). London: Wiley Interscience.

LaBerge, D. (1980). Unitization and automaticity in perception. In J. H. Flowers (Eds.), NEBRASKA SYMPOSIUM ON MOTIVATION (pp. 53- 71). Lincoln, NB: University of Nebraska Press.

Lindsay, P. H., and Norman, D.A. (1977). HUMAN INFORMATION PROCESSING: AN INTRODUCTION TO PSYCHOLOGY, SECOND EDITION. New York: Academic Press.

Mandler, G. (1984). MIND AND BODY: PSYCHOLOGY OF EMOTION AND STRESS. New York: Norton.

Mangan, B. (1993). Taking phenomenology seriously: The "fringe" and its implications for cognitive research. CONSCIOUSNESS AND
COGNITION, 2(2), 89-108.

Marcel, A. J. (1983). Conscious and unconscious perception: An approach to the relations between phenomenal experience and perceptual processes. COGNITIVE PSYCHOLOGY.

Milner, A. D., &;Rugg, M. D. (Ed.). (1992). THE NEUROPSYCHOLOGYOF CONSCIOUSNESS. London: Academic Press.

Neisser, U. (1967). COGNITIVE PSYCHOLOGY. New York: Appleton- Century-Crofts.

Newman, J., &;Baars, B. J. (1993). A neural attentional model for access to consciousness: A Global Workspace perspective. CONCEPTS IN
NEUROSCIENCE, 2(3).

Norman, D. A., &;Shallice, T. (1980). ATTENTION TO ACTION: WILLED AND AUTOMATIC CONTROL OF BEHAVIOUR (CHIP Report No. 99). University of California, San Diego.

Rumelhart, D. E., McClelland, J.E., and the PDP Research Group (1986). PARALLEL DISTRIBUTED PROCESSING: EXPLORATIONS IN THE MICCROSTRUCTURE OF COGNITION, VOL. 1: FOUNDATIONS. Cambridge MA: Bradford/MIT Press.

Schacter, D. L. (1990). Toward a cognitive neuropsychology of awareness: Implicit knowledge and anosognosia. JOURNAL OF CLINICAL AND EXPERIMENTAL NEUROPSYCHOLOGY, 12(1), 155-178.

Shallice, T. (1978). The dominant action system: An information processing approach to consciousness. In K. S. Pope & J. L. Singer (Eds.), THE STREAM OF CONSCIOUSNESS: SCIENTIFIC INVESTIGATIONS INTO THE FLOW OF EXPERIENCE. NY: Plenum.

Shallice, T. (1988). Information-processing models of consciousness: Possibilities and problems. In A. J. Marcel &;E. Bisiach (Eds.), CONSCIOUSNESS IN CONTEMPORARY SCIENCE (pp. 305-333). Oxford: Clarendon Press.

Singer, J. L. (1988). Sampling ongoing consciousness and emotional experience: Implications for health. In M. J. Horowitz (Eds.),
PSYCHODYNAMICS AND COGNITION (pp. 297-346). Chicago:
University of Chicago Press.

Recommended Readings

Baars, B. J. (1988). A COGNITIVE THEORY OF CONSCIOUSNESS. New York: Cambridge University Press.

Crick, F. (1994). THE ASTONISHING HYPOTHESIS: THE SCIENTIFIC SEARCH FOR THE SOUL. New York, Little, Brown.

Flanagan, O. (199x) CONSCIOUSNESS RECONSIDERED. Cambridge, MA: MIT Press.

James, W. (1893). PSYCHOLOGY: BRIEFER VERSION. Cambridge, MA: Harvard University Press.

Tables

Table 1. Perception/imagery contrasts
Table 2. Capability contrasts