Part VII. Conclusion.
‹ ‹
Chapter Eleven
A summary and some future directions.
11.0 Introduction
11.1 Overall review.
11.2 Review of models.
11.3 Review of some theoretical terms and definitions.
11.4 What are the necessary conditions for conscious
experience?
11.5 Some practical implications of GW theory.
11.6 Some topics not covered here.
11.7 Some philosophical implications: Information and the
mind-body problem.
11.8 Future directions for research and theory.
‹< ‹
11.0 Introduction
We have come to the end of a long journey. For the author,
there has been a continual sense of surprise at the fruitfulness
of a few fundamental ideas in understanding a topic that has long
been thought to be out of bounds for respectable psychologists.
Of course whatever progress we have made is not complete, nor is
it conclusive or beyond reasonable dispute. No doubt there are
gaps and errors in the story presented here. There does seem to
be a kind of coherence, an underlying simplicity --- so many
well-established observations fall into place easily, with only a
few basic ideas.
A vast amount of experimental and theoretical work is
needed to test and clarify the hypotheses developed in this book.
That is in the nature of a "framework" theory like this one. It
is in many ways an invitation for further work, rather than a
monument of indisputable fact.
We cannot claim for the global workspace approach the kind
of certainty that one finds in long-established scientific
theory. But theory, in the first instance, is not a matter of
eternal verities; it is first of all a tool for thinking. A good
theory may not be ultimately true. It should however aim for the
greatest possible empirical adequacy and fruitfulness at this
moment, given our limited evidence and imagination. It will be
quite satisfactory if the present approach helps to move things
in the right direction, as a "scaffolding" for further work, to
use Freud's term.
11.1 Overall review. (SHOW FIGURE WITH MINI-MODELS?)***
Our theoretical framework has really only a few entities:
îspecialized unconscious processorsï, a îglobal workspace,ï and
îcontextsï. Indeed, contexts are stable coalitions of specialized
processors that have over time gained privileged access to the
global workspace (4.xx). Thus contexts reduce to specialized
processors. Further, there are only a few processing principles:
îcompetitionï through the global workspace, which can be viewed as
lowering activation levels of global messages, and îcooperationï,
which raises those activation levels. "Lowering and raising
activation levels" is of course analogous to excitation and
inhibition in neurons. There is also local processing, within
unconscious specialized processors, which does not require the
global workspace; but this may also work by means of spreading
activation (cooperation) and inhibition (competition) (e.g.,
Rumelhart & McClelland, 1986). In sum, three entities and two
processing principles together can explain a vast range of‹j ‹
evidence about consciousness, volition, and the organization of
self.
11.2 Review of models.
Developing a theory is much like tracing a consistent path
through a maze of theoretical options. The maze may be quite
complex, and certainly the number of îpotentialï paths is huge, but
the most plausible path, given current knowledge, may be
relatively short and sweet. We can summarize the basic path we
have taken quite simply:
îFirstï we found a way to state empirical constraints that
must be met by any adequate theory of consciousness (1.xx).
îSecondï, this evidence suggested a system architecture in
which conscious contents are globally broadcast to a collection
of specialized unconscious processors (2.0). This is Model 1.
îThirdï, we explored the neurophysiological evidence, and
found it broadly consistent with this framework. However, there
is evidence not only for global broadcasting, but also for
feedback from the recipients of the global messages, and in
return to the message source as well.
îFourthï, we noticed that some stable sets of processors,
called contexts, must have privileged access to the global
workspace. Like other unconscious specialists, contexts can
compete and cooperate with each other to gain global access
(Model 2).
At this point the theory suggested that there were two
necessary conditions for conscious experience, namely global
broadcasting, and internal consistency --- the latter because
inconsistent global message inevitably trigger competition that
quickly destroys the global message.
îFifthï, in order to explain the fundamental general fact that
conscious experience fades with practice, we were forced to
postulate another major necessary condition for consciousness,
namely îinformativenessï; this was interpreted in the model as a
îchoice within a context of alternatives, demanding adaptation by
other processorsï. This created a functional role for the feedback
From the receiving processors to the global workspace, which w
already suggested by the neurophysiology (3.0). The whole system
now began to behave so as to seek a middle way between novelty
and redundancy; too much similarity, and conscious experience
will fade; too much mismatch with established context, and no‹j ‹
information can be interpreted (5.0, Model 3).
îSixthï --- without any further cost in theoretical entities,
the framework established so far led naturally to the notion of a
goal context or intention, which constrains conscious information
processing. This allows an explanation of "incubation" or
unconscious problem-solving phenomena, which are extremely common
--- ranging from word retrieval and the interpretation of
ambiguous stimuli to high-level artistic and scientific
achievement. The stream of consciousness can indeed be seen as a
flow of interacting conscious events and goal contexts (6.0,
Model 4).
îSeventhï --- the same ideas led to a modern interpretation of
William James' ideomotor theory of voluntary control, suggesting
that conscious goal images can by themselves trigger actions,
barring competing goal images or intentions. This in turn
suggested that the conscious component of voluntary action is
implicitly edited by multiple unconscious criteria (7.0, Model
5).
We were now able to explain a difficult ongoing puzzle,
namely the relationship between concrete qualitative conscious
events, such as we find in perception and imagery, and the
existence of non-qualitative "conscious" contents such as
concepts and immediate intentions to act. We suggested that non-
qualitative concepts may be under ideomotor control --- that is,
that there are in fact fleeting conscious images involved in
abstract concepts and intentions, but that these images are
difficult to retrieve. They do give rapid access to abstractions
that are not qualitative. Thus ultimately, all conscious events
involve qualitative phenomena, even though some may be quite
difficult to retrieve. This hypothesis needs empirical testing,
of course.
îEighthï. Within the GW framework it made sense to draw a
distinction between îconscious experienceï as such, vs. îattentionï
as the control of access to conscious experience. The traditional
ideas of voluntary and involuntary attention can then be
interpreted as access that is due to previous conscious events,
vs. access that is controlled unconsciously (8.0, Model 6).
îNinthï. Along the way, and much to the author's surprise, it
also became natural to interpret notions of "self" and "self-
concept" as the deeper parts of the context hierarchy (9.00,
Model 7). A set of empirical constraints contrasting self-
attributed vs. self-alien phenomena seemed consistent with this
approach.
îTenthï. Given the development so far, we were able to suggest
eighteen basic adaptive functions of consciousness. The most
fundamental one was the ability to support cooperative
interaction between multiple knowledge sources, so as to cope
with novelty. ‹j ‹å
A number of empirical predictions were made along the way.
Explanations emerged quite naturally for unexpected topics like
mental absorption, absent-mindedness, hypnosis, the existence of
self-alien symptoms in the clinic, involuntary slips, mental
effort, decision-making and indecisiveness, emotional upset as a
deep disruption of context, surprise as a momentary "erasure" of
the global workspace, and the like. There was a continual sense
of surprise that new domains of evidence appear to flow quite
spontaneously from the developing theory in the most natural way.
This repeated esthetic experience --- of surprising simplicity
underlying the apparent complexity --- remains for the author the
single best argument for the current approach.
11.3 Review of some theoretical terms and definitions.
The trickiest are: context, ideomotor flashes, self as I vs.
self-concept as me, the extent of global broadcasting,
informativeness, contextualization/objectification, decentralized
processing, Redundancy Effects, hypnosis, surprise. See Glossary.
11.4 What are the necessary conditions for conscious
experience?
We can now summarize five necessary conditions without
which conscious experience of an event is lost. They are as
follows:
îConscious events involve globally broadcast informationï.
This is quite a strong claim to make, but there is
considerable evidence in its favor (2.xx). Further, a number of
the theoretical claims made throughout this book are based on it.
For example, the ideomotor control of action would not work
unless conscious messages were made available to potentially all
effectors and action schemata (7.0). The notion of universal
editing of conscious goal images (7.xx) would not work unless any
editing criterion can compete against a globally broadcast goal
image. And so on.
îConscious events are internally consistent.ï
Again, the evidence for this idea from both perception and
cognition is quite good (x.xx). Chapter 2 argued that other
features like limited capacity and seriality follow from the
internal consistency constraint.
îConscious events are informative --- that is, they place a
demand for adaptation on other parts of the system.ï
Chapter 5 was devoted to pursuing this claim, and the
complementary hypothesis that conscious events that become
predictable fade from consciousness, though they do not disappear
--- in fact, faded conscious events may create the context for
later conscious events. These facts imply that consciousness
requires a global message to be available long enough for many
local processors to adapt to it.
îConscious events require some adaptation in the organization
of "self".ï
In a sense this claim is like the preceding one; conscious
events place a demand for adaptation on the system. The deeper
layers of context may be perceived to be "self-like" (Chapter 9),
in that strong violations of these deeper layers are experienced
as self-alien. These deeper layers may have to respond adaptively
to conscious events, either by generating a voluntary response to
do something about the event, or simply by recording that it has
happened. In that sense, the deeper layers of context function
much like conventional long-term memory --- they record
experienced events. (LTM ***?)
îConscious experience has a special affinity for perceptual
and quasi-perceptual events; even abstract conscious concepts may
involve rapid quasi-perceptual events.ï
The first arguments for this position were given in Chapter
1 (1.xx). Perception, imagery, bodily feelings and inner speech
seem to be involved in the conscious components of thought and
action, not merely in input processes. This suggests that
perception may be closer to the mind's îlingua francaï than other
codes. The evidence is good that percepts and images become
automatic with practice, and thus fade from consciousness.
Nevertheless, it seems likely that such faded quasi-perceptual‹j ‹
events continue to serve as a processing code, at least to evoke
previously well-prepared processors. For example, the evidence
gathered by Pani (1982) suggests as much (1.xx).
îWhat is unconscious?ï
If these are înecessaryï conditions for conscious experience,
it follows that anything that violates just one necessary
condition is unconscious. That is, events that are globally
broadcast but internally inconsistent are presumably not
conscious; perhaps they last for such a short time before their
competing parts destroy the global message, that they fail to
trigger an informative demand for adaptation. Similarly, it is
conceivable that contextual informtion could be globally
broadcast without being informative, because the system at large
has îalreadyï adapted to contextual constraints. There are thus
many ways in which information may be unconscious.
In sum, we find again that surprising simplicity emerges
From the apparent complexity. The evidence discussed througho
this book seems to converge on only five necessary conditions for
conscious events: global broadcasting, internal consistency,
informativeness, adaptation in the organization of self, and
perceptual or quasi-perceptual coding.
11.5 Some practical implications of GW theory.
Nothing, it is said, is as practical as a good theory. If
that is true, and if the present theory is reasonably good, it
should have some practical uses. Consider the issue of self-
control for example. Self-control is vital in psychopathology,
but also in children's learning to control socially undesirable
impulses, or in the desire of adults to control unwanted habits.
The ideomotor theory suggests that îifï one can be conscious of a
desired action without competition for a long enough time
(probably on the order of seconds) --- that the desired behavior
will take place, and undesirable behavior will be avoided. The
tricky question is, of course, how to avoid competing goal images
--- the images of eating compulsively, of smoking, of expressing
inappropriate hostility, etc. --- without becoming caught in an
active struggle for access to consciousness. GW theory suggests
that one way may be to use a strong context: something that
already structures a flow of conscious contents in the right way.
For instance, mnemonic techniques may be used to structure a flow
of constructive conscious images, lasting long enough to
eliminate an undesirable conscious goal. Or a dominant belief
system may create the proper context for action.
These suggestions are merely illustrative, of course. We‹j ‹
cannot explore these suggestions here in detail. No doubt these
very difficult practical problems will not be instantly solved
within the theory developed so far; but it may be able to provide
a useful framework for addressing them.
11.6 Topics not covered here.
This book has slighted certain topics that may yet be
addressed from the GW perspective. It is worth remembering what
we have not tried to explain. For instance, we have said little
about time perception, even though it is clearly related to the
flow of conscious thought (Blumenthal, 1977; Fraisse, 1963;
Ornstein, 1969). We do not at this point have a satisfactory
explanation of the "seven plus or minus two" size limit on Short
Term Memory. We have scarcely begun to address the question of
reflective thought, which is the main issue in much psychological
research (e.g. Newell & Simon, 1972). Except for our discussion
of hypnosis and absorbed states (x.xx), we have not dealt at all
with altered states of consciousness. All over the world there
seem to be universal similarities in mystical experiences,
spanning the length of recorded history (Huxley, 1970). In
principle, we should be able to say something useful about this
topic too. Finally, we have not really addressed the important
issues surrounding sleep and dreaming. If the present framework
proves to be viable, it will be worthwhile to explore the
implications of GW theory for such significant questions.
11.7 Philosophical implications: The mind-body problem
revisited.
Some philosophical questions may be so basic that they are
never wholly settled. Several scientists have argued that the
mind-body problem may be one of these, and that in fact "mind"
has emerged in modern science in the guise of "information,"
which plays a central role in physics, biology, and the
information sciences (e.g. Szilard, 1929/1964; Bakan, 1980;
Shannon & Weaver, 1949; Wiener, 1961; Baars, 1986 a). Information
is said to be irreducible to physical quantities. As Norbert
Wiener has written, "Information is information; it is neither
matter nor energy." (1961) If the claim that consciousness must
be informative can be developed (5.xx), we might add psychology
to the list of sciences for which the concept of information is
central. Further, information and control are two sides of the
same mathematical coin (Wiener, 1961), just as consciousness and
volition are two sides of the same psychological coin (7.0). We
can only mention these intriguing hints in passing, but they may‹j ‹
be well worth following up.
11.8 Future directions.
There is a great need to model the ideas developed in this
book more explicitly. This is not very difficult; in fact, there
have been simulations of "blackboard systems" since the 1970s
(see Chapter 2). We have begun to develop a simple simulation
intended to capture some central features of GW theory, using the
techniques of "connectionist" models (Cohen, Galin & Baars,
1986). But much more needs to be done.
Further, we need more critical experiments. One basic
prediction from GW theory was apparently confirmed even before
this book came to press (Greenwald, 19xx, 2.xx, 5.xx). But of
course a framework theory like this requires more than a single
critical experiment. We could easily imagine a comprehensive
research program to rigorously test the empirical predictions,
and to improve the precision and consistency of the theoretical
claims.
The work is never done, of course. Nevertheless, in glancing
back to the beginning of this book, we can take some satisfaction
in the thoughts that many difficult and confusing issues seem
clearer now, and that a great number of facts seem to fit this
simple framework surprisingly well. Whether this sense of clarity
and integration is real, or whether it is destined to disappear
with further findings and thinking, only time will tell.