Part IV. Goals and voluntary control.
So far we have considered what it means for something to be
conscious. In this section we place these considerations in a
larger framework, exploring the uses of consciousness. Thus we
move away from a consideration of separate conscious events îï to a
concern with conscious îaccessï, îproblem-solvingï and îcontrolï.
Chapter 6 describes the commonly observed "triad" of conscious
problem assignment, unconscious computation of routine problems,
and conscious display of solutions and subgoals. This triadic
pattern is observable in many psychological tasks, including
creative processes, mental arithmetic, language comprehension,
recall, and voluntary control. It suggests that conscious
contents often serve to assign problems to unconscious processors
which work out routine details, constrained by a goal context.
The interplay between conscious contents and goal contexts also
provides a plausible account of the stream of consciousness. In
Chapter 7, a contrastive analysis of voluntary versus involuntary
actions leads to a modern version of James' ideomotor theory,
suggesting that voluntary actions are also recruited by conscious
goal images. This view fits smoothly into our developing theory.
‹H ‹
Chapter Six
Model 4:
Goal contexts, spontaneous problem-solving
and the stream of consciousness.
"Every actually existing consciousness
seems ... to be a îfighter for endsï, of
which many, but for its presence, would
not be ends at all."
--- William James (1890)
6.0 Introduction.
6.1 The Tip©of©the©Tongue State as a îgoal contextï or îintentionï.
6.11 William James on the tip-of-the-tongue experience.
6.12 Theoretical implications of James' observations.
6.13 Intentions in GW theory.
6.2 The conscious-unconscious-conscious triad.
6.21 Model 4: Triadic patterns in the model.
6.22 Conscious contents can evoke new goal contexts.
6.23 Goal contexts can also evoke new conscious
contents.
6.24 The great generality of triadic problem-solving.
6.3 Empirical assessment of goal contexts.
6.4 Goal contexts and the stream of consciousness.
6.41 Competing and cooperating goal contexts.
6.42 The goal hierarchy represents significance.
6.43 General-purpose utility contexts.
6.5 Further implications.
6.51 Why it is important for goals to compete for access to
the global workspace.
6.52 An answer to the problem of non-qualitative conscious
events?
6.6 Summary.
‹j ‹å6.0 Introduction.
Consider the following questionsº º1:
(1) What are two names for the ancient flying reptiles?
(2) What technology develops artificial limbs and organs?
(3) What are three synonyms for "talkative"?
These questions evoke a mental search for words that are
known but rare. The search may take longer than expected, tending
to create what William James called a "tip-of-the-tongue" (TOT)
state, in which we have an intense feeling of knowing the word in
question, even though it does not come to mind immediately (Brown
& McNeill, 1966). This chapter is about this state in all its
variegated forms.
In the last two chapters we explored the îcontextsï of
experience, defined as those systems that shape and bound
conscious experiences without being conscious themselves. In this
chapter we show how the very general idea of a îgoal contextï or
îintentionï allows us to deal in a natural way with tasks that
extend over more than a single conscious experience. In practice,
of course, all psychological tasks involve more than a single
conscious event. The notion of a goal context allows us to
understand a very large set of phenomena as variants of a single
pattern. Creative processes in art, science and mathematics seem
to be under the control of goal contexts --- but so are short-
term events like word-search, question-answering, the
interpretation of ambiguous words and figures, control of action,
and the like (6.24). The stream of consciousness can be
considered as a flow of experiences created by the interplay of
many goal contexts, each tending to make conscious whatever will
promote progress toward its goal (6.4).
Notice that goal contexts are not necessarily îlabeledï as
voluntary. The ability to label one's own goals and to guide
one's own processes requires an additional layer of metacognitive
organization, to be discussed in Chapters 7, 8 and 9.
We begin with William James' well-known description of the
intention to remember a forgotten word --- the tip-of-the-tongue
(TOT) state --- and conclude, in modern terms, that the TOT state
is a complex representational state that takes up limited
capacity, that guides word-search and evaluates candidate words,
but that does înotï have experienced qualities like color, warmth,
flavor, pitch, and a clearly bounded locus in time and space.
This state differs therefore from mental images, inner speech,
feelings, or percepts, which do have experienced qualities (cf.
Natsoulas, 1982). James suggests that perhaps one-third of our
psychic life may be spent in such states of specific expectation.
We will pursue the argument that the TOT state represents "a goal
context searching for a conscious content." ‹j ‹å
Such goal contexts are different from conscious events that
function as goals; both are needed for the system to operate.
Given a conscious event that can be interpreted as a goal, the
system works to recruit a goal context, and a set of processors
that engage in "spontaneous problem-solving." The resulting
solutions often become conscious. If the goal cannot be reached,
obstacles and subgoals tend to become conscious, recruit their
own resources, until finally an acceptable conscious solution
emerges.
This kind of spontaneous problem-solving is extremely
general. It can cover life plans, fantasies for the future,
control of one's own body, retrieval of the right word at the
right time, sentence comprehension, attempts to achieve social
influence, and an endless variety of other goals that people
think of achieving at one time or another. Even operant and
classical conditioning can be seen as types of goal-directed
problem-solving. The material in this chapter is therefore
crucial to the claim that conscious processes are îfunctionalï:
that they help people achieve their goals in life.
6.1 The Tip©of©the©Tongue State as a îgoal contextï or
îintention.ï
6.11 William James on the tip-of-the-tongue experience.
We begin with the following observations from William James
about the state of attempting to recall a forgotten word (1890).
Is such a state truly conscious or not, asks James?
"Suppose we try to recall a forgotten name. The state of our
consciousness is peculiar. There is a gap therein; but no mere
gap. It is a gap that is intensely active. A sort of a wraith of
the name is in it, beckoning us in a given direction, making us
at moments tingle with the sense of our closeness, and then
letting us sink back without the longed-for term. If wrong names
are proposed to us, this singularly definite gap acts immediately
so as to negate them. They do not fit into its mould. ...
Thus clearly something is going on --- we are conscious of
some sort of îdefiniteï state, because if someone suggests the
wrong word to us, we know immediately that this îisï the wrong
word. And we also immediately recognize the right word when it
comes to mind. In modern terms, we can successfully recognize
îmatches and mismatchesï of the state of looking for a forgotten‹j ‹
word --- and the ability to accurately detect matches and
mismatches implies that this state involves a îrepresentationï of a
target word. Since words can vary along many dimensions, it must
be a îcomplexï representational state, much like a mental image or
a percept.
Further, this "tip-of-the-tongue" state resembles a mental
image or a percept, because having it excludes other conscious
contents. We cannot search for a forgotten word and at the same
time contemplate a picture, think of yesterday's breakfast, or do
anything else that involves conscious experience or mental
effort. The TOT state occupies our central limited capacity.
But in one respect the TOT state differs from mental images,
feelings, inner speech, and perceptual experiences. All these
conscious events have experienced qualitative properties ---
qualities like size, color, warmth, or location. But the TOT
state does not have experienced qualities (viz. Natsoulas, 1982).
Two different TOT states are înot experiencedï as sounding
different, even though the words they stand for sound different.
In some ways, therefore, the TOT state is like other conscious
states like percepts and images; in other ways, it is not like
those conscious experiences at all, but much more like the
contexts discussed in Chapter. 4.
The same may be said whenever we intend to speak a thought
that is not yet clothed in words:
"And has the reader never asked himself what kind of a
mental fact is his îintention of saying a thingï before he has said
it? It is an entirely definite intention, distinct from all other
intentions, an absolutely distinct state of consciousness,
therefore; and yet how much of it consists of definite sensorial
images, either of words or things? Hardly anything! Linger, and
the words and things come into the mind; the anticipatory
intention, the divination is there no more. But as the words that
replace it arrive, it welcomes them successively and calls them
right if they agree with it, it rejects them and calls them wrong
if they do not. It has therefore a nature of its own of the most
positive sort, and yet what can we say about it without using
words that belong to the later mental facts that replace it? The
intention îto-say-so-and-soï is the only name it can receive."
James suggests that perhaps one-third of our psychic life
consists of states like this; further, he seems to say that this
state itself triggers off retrieval processes, which produce the
words that will clothe the intention. In other words, the TOT
state is active; it initiates a conscious display of a series of
candidate words, and "it welcomes them ... and calls them right
if they agree with it, it rejects them and calls them wrong if
they do not."
‹j ‹å
6.12 Theoretical implications of James' observations.
We can summarize in modern terms the theoretical claims made
by James about the tip-of-the-tongue state:
1. The TOT state involves a complex representation of the
missing word. (As shown by the fact that it accurately matches
and mismatches candidate words.)
2. The TOT state occupies central limited capacity, like
other conscious states. (Witness the fact that the TOT state
excludes mismatching candidate words.)
3. The TOT state helps trigger off word retrieval processes,
so that candidate words come to consciousness as long as this
state dominates our limited capacity.
4. The TOT state only stops dominating our central limited
capacity when the right word is found.
6. And yet in spite of all these properties the TOT state
does înotï have experiential qualities like color, warmth, flavor,
location, intensity, etc. It is therefore radically different
From other conscious experiences like mental images, feeling
inner speech, and percepts (Footnote 1).
These observations apply generally to intentions and
expectations. To create experiences like this for îanyï action, we
need only ask someone to perform the action and then delay the
moment of execution. To have a runner experience a "tip-of-the-
foot" experience, we need only say "GET READY", "GET SET", and
then delay "GO". At that point the runner is poised to go, the
"intention" is at its highest pitch, and yet the action is not
executed. There may be no sensory experience of the "intention to
run," but the runner's concentration will still be impaired by
interfering conscious events. The "intention to run" takes up
limited capacity just like the tip-of-the-tongue state does.
Given these implications, let us sketch out a way in which
the global workspace theory can model the TOT state, its
antecedents and its consequences.
‹j ‹å 6.13 Intentions in GW theory.
GW theory treats the TOT state as a îcurrent goal contextï, an
unconscious structure that dominates our limited capacity for
some time --- witness the fact that it competes with any
conscious content. But the current goal context naturally
operates in its own higher-order, more permanent context of pre-
existing goals (4.xx). At this moment the reader's conscious
experience is presumably shaped by the goal of reading this
sentence. But as soon as the sentence ends, or the book is
closed, it becomes clear that this local goal context exists
always in its own complex hierarchy of goals, in which reading
the sentence is merely a local subgoal.
The claim made here is that goal contexts are the same as
the îintentionsï of common sense psychology. We will use these
expressions interchangeably. The term "goal context" emphasizes
the contextual and non-qualitative nature of intentions, and
their similarity to other contexts, especially conceptual
contexts (4.xx). Indeed, we can say that conceptual contexts are
equivalent to expectations, and goal contexts are equivalent to
intentions. But intentions and expectations are very similar:
indeed, one can argue that they are basically the same. We speak
of expectations when we have a future-directed mental
representation that is dependent on external events for its
satisfaction. An intention is the same, except that it depends
on internal events. An intention, then, is an expectation about
oneself.
A goal context does not have to be evoked by verbal
questions. Any conscious stimulus that is incomplete, that
invites further processing, seems to initiate a context that
guides further unconscious work. This has been widely noted, for
example by Gestalt psychologists, by the Wuô"rtzburg school, by
Zeigarnik and by Ach (Rapaport, 1955; Murray, 1983). It is most
widely known as the "Zeigarnik phenomenon," the tendency to
complete incomplete mental contents. There has been some
controversy about the evidence for it recently, which we will
discuss below (Holmes, 1967, 1968).
If it is generally true that intentions do not have
qualitative conscious contents, then a current controversy about
the ability to report intentions begins to make more sense.
Nisbett and Wilson (1977) cite a number of social psychological
studies showing that the intentions people attribute to
themselves can often be quite incorrect. For example, in a
department store display of socks or perfume, people will tend
statistically to choose the right©most item; yet if asked why
they chose the item, they will produce all sorts of reasons other
than the one that seems to be operative. This is only one of
dozens of demonstrations showing that people have poor access to
their own reasons for doing things, even leaving out those cases
where the rationalization is self©serving or defensive. But we‹j ‹
also know that under optimal conditions people can report other
mental processes quite accurately (e.g. in mental imagery,
explicit verbal problem solving, rehearsal in short-term memory,
etc.; see Ericsson & Simon, 1980). What is the difference then
between reporting intentions and reporting mental images? Why are
intentions so difficult to report?
One possible explanation is that intentions are complex,
înon-qualitativeï but capacity-limiting events, which are not
experienced in detail. To become reportable, intentions must be
converted into an introspectible code, such as inner speech,
visual images, or perhaps bodily feelings. These conscious
contents may be easier to report accurately. This is not to say
that intentions have no conscious correlates: there may be
qualitative images, inner speech, etc., associated with the
intention. Further, when the intention runs into difficulties
there is a conscious sense of surprise. But such conscious
contents are înot the same as the intention itselfï.
Below we will attempt to specify the role of consciously
experienced goals, as important parts of any intention (see also
Chapter 7). But first, we will specify the notion of intention or
goal context in more detail.
îIntentions as multi-leveled goal structuresï
Intentions or goal-contexts represent future states of the
system, serving to constrain processes that can reach those
future states. But intentions are not simple goals. They must be
multi-leveled goal îstructuresï, consisting of numerous nested
goals and subgoals. Even a single spoken sentence is constrained
by many simultaneous goals and criteria, including those that
specify the desired loudness and rate of speech; voice-quality;
choice of words; intonation; dialect; morphology; syntax; choice
of rhetorical style; semantics; discourse relations;
conversational norms; and communicative effectiveness (Clark &
Clark, 1977). Each of these levels of organization can be
described in terms of general goals, which the action can match
or mismatch. Each of these levels can go astray, and errors in
each level of control can be detected and corrected immediately.
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
Insert Figure 6.13 about here.
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
On top of these linguistic criteria we use language to gain
a multitude of pragmatic ends, many of which combine to constrain
any single speech act. Thus we may routinely want to appear
educated in our speech, but not stuffy; tolerant, but not
undiscriminating; we may want to capture the listener's
attention, but not to the point of screaming for it. All such‹j ‹
pragmatic goals simultaneously constrain any speech act.
Notice again, that contexts can compete for limited capacity
just as conscious contents can. But there is this difference,
that we can only be aware of one chair (one conscious content),
but that many goal contexts can simultaneously constrain limited
capacity as long as they are mutually compatible (Figure 6.2).
Once established, linguistic and pragmatic goal systems do
not become conscious as a whole. Thus at the minimum, an
"intention to say something" must involve a îmany-leveled goal-
structureï, in which each major goal can activate numerous
subgoals to accomplish its ends. At any one time, most components
of such goal structures are not qualitatively conscious. Figure
6.2x represents such goal structures as graphic horizontal
"frames", which together constrain any action.
The key observation is that goal contexts are apparently
triggered by conscious events, and that they result in other
conscious events. This îtriadic patternï appears to be of very
great generality. We explore it next, and then modify our model
to accomodate it.
6.2 The conscious-unconscious-conscious (CUC) triad.
The common pattern here is:
1. a conscious or expressible îstage of problem-assignmentï;
2. an unconscious îstage of processingï guided by an
unconscious, limited-capacity-loading goal
context;
3. a conscious îdisplay of the solutionï.
In the case of complex problems, these three stages can
contain subgoals with their own triads (Figure xx). For example,
mental arithmetic might work roughly as follows:
1. CS goal-assignment: "multiply 12 x 24."
2. UCS search for subgoals.
3. CS subgoal assignment: "first multiply unit values by
24."
4. CS subgoal assignment: "multiply 2 x 24."
5. UCS search for answer.
6. CS report of subgoal: "... = 48."
7. CS subgoal assignment: "now multiply 10 x 24."
8. UCS processing: ("add a zero to multiply by
10").‹j ‹å
7. CS display of subgoals: "48, 240."
8. CS subgoal assignment: "now add the two outcomes."
9. UCS addition process: ("48 + 240").
10. CS display of solution: "... = 288."
This example is only illustrative. There must be much
individual variation. A very skilled mental calculator might only
be conscious of steps (1) and (10), all others being automatic
and unconscious. A novice might need more conscious intermediate
steps than are shown here. The point is that a triad with this
conscious-unconscious-conscious (CUC) pattern can be expanded
into a series of triadic cycles, with as many cycles as there are
subgoals to be satisfied. Table 6.2 shows this as a contrastive
pattern.
---------------------------------------------------------------
Table 6.2
Triadic problem-solving pattern.
---------------------------------------------------------------
îSimple Case:ï
1a. Conscious problem assignment. b. Unconscious processing
c. Conscious solution appears.
---------------------------------------------------------------
îMore general case:ï
1a. Conscious problem assignment b. Unconscious
comprehension
of problem parameters
(goal context)
c. Unconscious processes
within goal context
d. Conscious sub-goal solution.
2a. Conscious assigment of further
sub-goals
b. Add goal context
c. More unconscious
problem-solving
(Repetition of loop 1a - 1d until solution is reached.)
d. Conscious solution appears.
----------------------------------------------------------------¨‹j ‹å
6.21 Model 4: CUC triads in the model.
It is easy to incorporate these ideas into our theoretical
diagrams. We need only show the global workspace over time, and
indicate that global messages broadcast both to potential goal
contexts and to unconscious processors able to carry out the
goals. Thus goal contexts can be recruited, and when they are,
they compete for access to the global workspace. Once a give goal
context becomes dominant, it begins to limit the conscious events
that are likely to gain GW access. Thus an intention to retrieve
the name of the ancient flying reptiles will restrict conscious
contents until the right word comes to mind (6.0). However, as
pointed out above, this intention must itself exist in a
hierarchy of other intentions, so that, if the word remains out
of reach, at some point a higher-level goal structure may decide
that the game is not worth the candle, and that one had better
give up, wait, or look at a dictionary. Alternatively, other
local goals can start competing against the intention to retrieve
the word. In any case, local intentions are not absolutely
dominant.
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
Insert Figure 6.21 about here.
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
The triadic pattern results naturally from this model.
Conscious events, themselves constrained by higher-level goal
contexts, can recruit a local intention. In its turn, the local
intention allows a mental image representing a problem to become
conscious. This serves to recruit processors and subgoals able to
solve the problem. If these can operate automatically, they
simply compute a solution to the problem. If they reach a dead
end, they may be able to recruit their own conscious subgoal
image, to broadcast the obstacle, thereby mobilizing additional
processors able to solve the resistant sub-problem, etc., etc.,
until finally the original goal is solved. Alternatively, the
original goal context may be competed out of dominance of the
global workspace --- in that case, we simply give up on the
problem for the time being.
Notice that there is no îself-consciousï problem-solving in
this model. That is, the system does not know metacognitively
that it is solving a problem. It simply does what it does,
"spontaneously." We will come to self-conscious metacognition in
a later chapter (8 and 9).
‹j ‹å 6.22 Conscious contents can trigger new goal contexts.
In language, beginning a new topic or asking a question
serve to access a goal context. Asking a question creates a set
of constraints on subsequent conscious events: for at least a few
seconds, conscious thoughts are constrained to be responsive to
the question (Miyake & Norman, 1978). This is indeed how Brown
and McNeill (1966) elicited TOT experiences experimentally, by
asking subjects to find a fairly rare word corresponding to a
definition (6.0). Again, the question is conscious, the
constraints on answering are not conscious in detail, nor are the
routine details searching for the answe; but the answer, when it
comes, does become conscious.
Conscious events that are experienced as îincompleteï seem to
set up very active goal contexts. They may vary from incomplete
sentences or visual figures, to a social or emotional issue, to
an unanswered mathematical or scientific problem. In each case,
the conscious experience seems to set up unconscious constraints
on future conscious events. This effect was demonstrated directly
by Klos and Singer (1981) in the case of unresolved emotional
conflicts in adolescents. For college students with persistent
parental conflicts, a conflict-situation was enacted
dramatically; for example, there might be an argument about
borrowing the key to the parents' car. Then the students were
asked to lie down in a quiet room to report their spontaneous
thoughts. When the dramatic re-enactment was unresolved, there
were significantly more thoughts about it than when it was
resolved (viz. Singer, 1984). The great advantage of this study
is its human relevance; it may be more difficult to obtain in
laboratory tasks that are perceived to be irrelevant to the
subjects' everyday lives, but this should not be surprising
(Holmes, 1968).
6.23 Goal contexts can also evoke new conscious contents.
If the TOT state is indeed "a goal context looking for the
right content," it already provides us with one example of a
dominant context triggering new conscious contents. The "Aha!"
experience in problem-solving is another example, as is the
conscious "popping up" of answers to questions, the emergence of
thoughts in free association, and the like. Thus it works both
ways; conscious events can evoke goal contexts, and these can in
turn evoke new conscious thoughts and images. This completes the
conscious-unconscious-conscious triad.
‹j ‹å
6.24 The generality of triadic problem-solving.
The triadic pattern is extremely common. We find it in
explicit problem-solving, but also in day-dreaming (Singer, 1984;
Klinger, 1971); it is the common pattern in controlling even
simple actions, but it can also be found in long-term planning;
it appears in perception, for example in the well-known bi-stable
figures (the Necker cube, figure-ground illusions, etc.), as well
as in memory retrieval, as shown by the Tip-of-the-Tongue
phenomenon. Finally, and most well-known of course, it appears in
high-level creative work. We will now examine a sampling of this
general phenomenon, starting with examples from art, science, and
mathematics --- the highest levels of human creativity (Hadamard,
1945; Ghiselin, 1952; John-Steiner, 1986).
A. îHigh-level creativityï
The role of unconscious problem-solving in high-level
creativity was described early on by the mathematician Henri
Poincar'ôe, who devoted much thought to the psychology of
mathematical creation. He wrote,
"Most striking at first is this appearance of sudden
illumination, a manifest sign of long, unconscious prior work.
The role of this unconscious work in mathematical invention
appears to me incontestable, and traces of it would be found in
other cases where it is less evident. Often when one works at a
hard question, nothing good is accomplished at the first attack.
Then one takes a rest, longer or shorter, and sits down anew to
the work. During the first half- hour, as before, nothing is
found, and then all of a sudden the decisive idea presents itself
to the mind. ... There is another remark to be made about the
conditions of this unconscious work: it is possible, and of a
certainly it is only fruitful, if it is on the one hand preceded
and on the other hand followed by a period of conscious work".
(Ghiselin, 1952, p. 38).
This last remark of course describes what we have called the
"conscious-unconscious-conscious" triad. This is also emphasized
in the following quote from the poet Amy Lowell (Ghiselin, 1952
p. 110):
"How carefully and precisely the subconscious mind
functions, I have often been a witness to in my own work. An idea
will come into my head for no apparent reason; 'The Bronze
Horses,' for instance. I registered horses as a good subject for
a poem; and, having so registered them, I consciously thought no
more about the matter. But what I had really done was to drop my‹j ‹
subject into the subconscious, much as one drops a letter into
the mailbox. Six months later, the words of the poem began to
come into my head, the poem --- to use my private vocabulary ---
was 'there'". (Ghiselin, 1952, p. 110).
Of course, creative people are often conscious of
intermediate events in this process, which we would interpret as
subgoal processing (6.2x). And of course not all creative work is
experienced as spontaneous --- some of it is effortful and
deliberate. This mixture of ingredients goes to make up a
completed work. Listen to Mozart:
"When I am, as it were, completely myself, entirely alone,
and of good cheer ... my ideas flow best and most abundantly.
îWhenceï and îhowï they come, I know not; nor can I force them. Those
ideas that please me I retain in memory ... If I continue in this
way, it soon occurs to me how I may turn this or that morsel to
account, so as to make a good dish of it, that is to say,
agreeably to the rules of counterpoint, to the peculiarities of
the various instruments, etc.
"All this fires my soul, and, provided I am not disturbed,
my subject enlarges itself, becomes methodised and defined, and
the whole, though it be long, stands almost complete and finished
in my mind, so that I can survey it, like a fine picture or a
beautiful statue, at a glance. Nor do I hear in my imagination
the parts îsuccessivelyï, but I hear them, as it were, all at once
(îgleich alles zusammenï). What a delight this is I cannot tell!
... When I proceed to write down my ideas, I take out of the bag
of my memory, if I may use that phrase, what has been previously
collected into it in the way I have mentioned". (Ghiselin, 1952;
p. 44).
It is clear that a major work is not accomplished in a
single conscious-unconscious-conscious leap. In fact, Poincareô'
may simply have forgotten some intermediate events between the
first effortful period of conscious problem-assignment and the
Aha! experience. Most problem-solving requires a string of CUC
triads.
So much for truly great creativity; we move now from the
sublime --- not to the ridiculous --- but to the commonplace, and
we find the same triadic pattern in "simple" events like
answering questions, retrieving memories, generating images,
switching between the two interpretations of an ambiguous event,
understanding analogies, generating free associations, and the
like.
‹j ‹åB. îDay-dreaming involves spontaneous problem-solving.ï
Thought-sampling studies indicate that for most people, a
substantial percentage of conscious activity does not serve a
self-conscious purpose. Several studies by Singer and his
colleagues, and by Klinger (1971) suggest that these
"daydreaming" activities may be quite goal-directed, even though
people may not be able to state their purposes. As Singer (1984)
has written (p. 25):
"...'current concerns' --- unfulfilled intentions,
hierarchically organized in terms of closeness to fulfillment or
personal value ... --- make up a goodly share of the conscious
content derived from thought sampling. Our basic 'rules' for
admission of material to consciousness seem to involve a
screening strategy that is especially sensitive to our current
concerns even in dichotic listening experiments or during sleep."
"...As we gain more information of this kind, it is my
personal guess that we will find that a great deal of respondent
(spontaneous) or playful fantasy has long-term planning,
decision-making, and self-schema formation functions."
C. îAnalogy tasksï
The following "remote associates" test devised by Mednick
(1962) is a good example of a large class of analogy tasks. The
task is to find the missing word. For our purposes it is useful
to allow the answer to come by itself, without deliberate effort.
For example, the following three words suggest a fourth:
1. cookies sixteen heart _____________
The answer "sweet" fits with cookies, with the phrase "sweet
sixteen", and with the word "sweetheart". Here are some more
examples º2:
2. poke go molasses _____________
3. surprise line birthday _____________
4. base snow dance _____________
5. elephant lapse vivid _____________
6. lick sprinkle mines _____________
7. stalk trainer king _____________
‹j ‹å On at least one of these items the reader should have a
genuine "Aha!" experience. Note that we sometimes feel great
certainty about the answer without knowing why.
D. îThe "magical" quality of learning and retrieval.ï
In Chapter 5 we maintained that most learning has this same
"magical" character. We simply pay attention to some material,
and learning seems to take place with no detailed self-conscious
guidance. Most people do not have a set of recallable rules by
which they learn.
Similarly, memory retrieval is typically un©self-conscious.
In memory tasks we often ask people to recall material
deliberately. But in speaking, in walking about the world, and in
performing a skilled action like driving a car, we retrieve
information from memory with little self-conscious recall. Most
retrieval is "magical" in the same sense that learning is.
E. îBi-stable figures in perceptionï
We can find the conscious-unconscious-conscious (CUC) triad
in bi-stable perceptual figures as well. We are conscious of one
interpretation of the Necker cube until it changes unconsciously,
and then we become conscious of the alternative intepretation.
The intermediate stage is unconscious. Much the same is true of
ambiguous words and sentences (Ch. 4). And the same pattern may
be found in "hidden figures" like the Dalmatian in the Park
(4.xx), which are bi-stable but not reversible. We are conscious
initially of the black-and-white blotches in the Dalmatian
picture, and we are conscious of the final stage, where we can
perceive the dog, the tree, and the sidewalk. But we are not
conscious in detail of the intervening stage of analyzing the
visual input to arrive at the new conscious intepretation.
F. îAction controlï.
Action control has much the same character, as we can see
From the extreme case of biofeedback training. When people lea
a biofeedback task such as controlling alpha waves in the
cortical EEG, they are not conscious of the îwayï in which they
control the alpha waves. They are conscious, in some broad sense,‹j ‹
of wishing to control a conscious feedback signal, but the
intermediate steps are simply not available to awareness.
However, the feedback signal itself is always conscious. Much the
same is true for any motor task, such as wiggling one's finger.
The intention to wiggle the finger has conscious or at least
expressible aspects, and we are conscious of wiggling the finger;
but the intermediate stage is unconscious. Most people do not
know that the muscles that move the finger are actually located
not in the hand but in the forearm. But they do not need to know
this --- it is part of the automatic, unconscious problem-solving
stage (x.xx).
6.3 Empirically assessing the existence of goal contexts.
A goal is a representation of a future state that tends to
remain constant as different means are explored for achieving the
goal. Like any other context, a goal context serves to bias
choice-points in processing. Perceptual contexts force the
interpretation of the Necker Cube, or the moon's craters, in one
direction rather than another (x.xx). Conceptual contexts work to
interpret the word "case" as "briefcase" rather than "sad case."
And goal contexts presumably cause different pragmatic
interpretations of ambiguous information, as well as different
choices in the control of action. These differences should then
allow us to assess goal contexts by presenting subjects with the
opportunity to interpret or act upon information that is
ambiguous with respect to the goal (Baars, 1985).
Work done on experimentally elicited slips of the tongue
provides a case in point (Baars, 1985; in press). A number of
techniques are now available for eliciting predictable slips of
the tongue. All these techniques create goal contexts in one way
or another. For example, one can ask people to repeat the word
"poke" over and over again. Since this is an action, it
presumably involves a goal context. When the subjects are then
asked, "What do you call the white of an egg?" they will tend to
say "the yolk," even when they know perfectly well this is the
wrong answer. Thus priming by repeating a similar©sounding word
works to structure later motor control and memory retrieval,
presumably by altering the goal context.
Slips may also reflect higher©level goal contexts. In îThe
ïîïîPsychopathology of Everyday Life ï(1901/1938, p. 81) Freud gives
the example of "a young physician who timidly and reverently
introduced himself to the celebrated Virchow with the following
words: 'I am Dr. Virchow.'The surprised professor turned to him
and asked, 'Is your name also Virchow?'" Slips like this are easy
to produce in the laboratory; one only needs to create confusion‹j ‹
between two alternative sentences, such as (1) 'Professor
Virchow, I am Dr. X' and (2) 'I am Dr. X, Professor Virchow.'
Competing sentences like these will result in inadvertent blends
such as 'I am Dr. Virchow, Professor X.' (Baars, 1980; 1985; in
press). Of course Freud argues that the slip reveals the
ambitious îpurposeï of the person making it. The young physician
wishes to be as famous as the great Virchow. In our theoretical
terms, Freud believed that the slip reveals a goal context.
This hypothesis may be somewhat more difficult to test, but
not impossible. It has been shown, for example, that male
undergraduates confronted with an attractive female experimenter
are more likely to make sexual slips of the tongue; when they
told that they may receive an electric shock during the
experiment, they make more shock©related slips (Motley, Camden, &
Baars, 1979). Thus slips seem to reflect important or immediate
goals and preoccupations. It is difficult to separate goal
contexts from related conceptual contexts that may be equally
primed. Surely when we are confronted with an intensely desired
object, our knowledge about it must also be activated. We should
not expect to find evidence for goal contexts without related
conceptual contexts. However, we would expect an increase in
slips that actually express a wish when there is a goal vs.
merely a conceptual preoccupation. Thus when we are hungry and
walk to the refigerator, we should expect this to prime goal™related
slips compared to the condition where we are not hungry,
and see someone else walk to the refrigerator. In both cases
conceptual contexts should be primed; but in the first, specific
actions related to the goals should be more likely as well.
6.4 Goal contexts and the stream of consciousness.
The stream of consciousness can be seen as a complex
interplay between conscious events and their goal contexts. Each
conscious event can trigger a new goal context, which can, in its
turn, evoke later conscious experiences (Figure 6.4). We
introduce a graphic notation for contexts. Competition between
incompatible contexts can cause a surprising "resetting" of
conscious contents. As we have noted before (x.xx) surprise may
lead to momentary forgetting and a resetting of the global
workspace due to competition between incompatible contexts
(Baars, 1987). A TOT state may result, which can be viewed as a
new, dominant context looking for a new set of conscious
contents.
©©©©©©©©©©©©©©©©-------------‹j ‹å
Insert Figure 6.4 about here.
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
6.41 Cooperating and competing goal contexts.
Some goal contexts compete against each other for access to
consciousness, but many goals must be compatible with each other
(4.xx). For instance, even a single sentence is constrained by
many simultaneous goals and criteria: vocal, linguistic,
semantic, and pragmatic. In Figure 6.4, nesting of contexts is
intended to symbolize that the contexts involved are compatible,
and indeed may be mutually supportive, while competing contexts
are shown operating at the same horizontal level.
We have some direct evidence for the operation of compatible
goal contexts from our experimental studies of unintended îdouble
entendresï (Motley, Camden & Baars, 1982). Double entendres, of
course, are compatible with two contexts of interpretation. Thus,
if subjects are induced to feel angry by post©hypnotic
suggestion, (context A), and if they are to select the best words
to fill out non©emotional multiple©choice items (context B), they
will tend to select items that are both angry and consistent with
the overt context. Thus Baars, Cohen and Bower (1986) presented
sentences like:
(1) I always lost at cards with him and wanted to .... more
often.
a. beat him (*); b. win; c. succeed; come off well.
(2) Toward the end of the day at the store, I still had a
few customers to ...... .
a. finish off (*); b. help; c. attend to; d. handle.
Angry subjects selected far more starred (*), ambiguously
angry items than happy subjects.
6.42 The goal hierarchy represents significance.
‹j ‹å
One of the great gaps in current cognitive theory is the
lack of a widely accepted account of significance or value. Some
things are more important to people than others: more
pleasurable, more painful, more valued, more likely to result in
action. Contemporary cognitive theory thus far has not attempted
to include this obvious and necessary dimension, but of course
there is overwhelming evidence that significance matters. The
great animal conditioning literature stands as a monument to this
truism, as does the vast clinical and social psychological
literature on emotion and motivation.
As pointed out in 4.xx, the most obvious way to incorporate
significance in GW theory is to use the hierarchy of goal
contexts (Figure 6.42). Some goals are more important than
others, and some provide the contextual presuppositions within
which other goals are defined. Thus the need to survive may be
presupposed in the need to eat, to escape from danger, etc. The
need to eat is presuppositional to the search for food. There is
a partial ordering between different levels of the goal
hierarhcy. But it is not invariable. The primacy of food declines
after eating and rises after food deprivation, so that the goal
hierarchy is re©ordered.
Significant events usually drive less significant ones from
consciousness (4.xx; 5.xx; 8.xx). We will discuss these cases
Chapter 8 under the heading of îprioritizingï of conscious
contents.
6.43 General-purpose goal contexts.
Some goal contexts must be useful in many situations. All
actions require control of timing, for example. Learning may
require organizational skills that are useful across many
different cases. It seems likely that there are îgeneral©purpose
goal contextsï, which can be called upon as subgoals in many
different tasks. Presumably these can be recruited just like any
other specialized processor. When a relevant message is
broadcast, they can compete for GW access. If they gain access,
they can become part of the current goal hierarchy in the same
way that any subgoal can become part of the current hierarchy
(Figure 6.21). Chapter 8 will propose that there are special goal
contexts in attentional control called options contexts. These
are presumed to help select conscious contents according to their
current significance.
‹j ‹å
6.5 Implications.
6.51 Why it is important for goals to compete for access to
the global workspace.
Several theories have been advanced with a similar notion of
action systems competing for access to a limited-capacity channel
(e.g. Shallice (1976) and Norman & Shallice (1981)). However,
these theories do not appear to explain the advantage of doing
so: why would action control systems bother to compete for such
access? GW theory gives one possible anser: If some goal system
needs to recruit other processors and subgoals to carry out its
goal, it must be able to dominate the global workspace long
enough to broadcast a "recruiting message" for the appropriate
specialists. In GW theory, access to the global workspace is the
key element for the recruitment of any new configuration of
systems. (We develop this idea further in the next chapter.)
Further, competition for access to consciousness is one way
in which dominant systems can also drive out or "repress"
disturbing or erroneous goals. Thus GW theory very naturally
allows us to model psychodynamic notions such as emotional
conflict and repression in terms of competition between different
goal systems for access to the global workspace. We have already
noted the idea that îsurpriseï may involve a momentary erasure of
the the GW, allowing new information to gain access to
consciousness (Underwood, 1982; Grossberg, 1982).
6.52 An answer to the problem of non-qualitative conscious
events?
One of our running themes involves the relationship between
conscious experience of qualitative events, such as perception
and imagery, îversusï conscious access to non-qualitative events,
such as currently conscious concepts, beliefs, etc. (1.xx, 2.xx)
One interesting possibility is that all abstract concepts are
accessed consciously by means of perceptual and imaginal events.
That is to say, it is possible that even the most abstract
concepts have qualitative mental symbols of some sort. This
hypothesis was popular about the turn of the century among
psychologists who were impressed by the fragmentary and fleeting
mental images that are often come with abstract concepts
(Woodworth, 1915). It has been advanced quite often by highly
creative people like Einstein and Mozart, in writing about their
own creative processes (x.xx; John-Steiner, 1986). We know of
course that the "imageless thought controversy" that came along
with this idea about the turn of the century was apparently not‹j ‹
very fruitful. But that does not mean it was wrong. In Chapter 7
we will propose that many concepts may be triggered by fleeting
conscious images. This ideomotor theory suggests a principled
connection between abstract non-qualitative concepts and
concrete, imageable conscious experiences.
It is attractive to simply suppose that non-qualitative
concepts are part of the contextual structure mentioned above.
Concepts can be evoked by qualitative events like percepts, inner
speech, and mental images; and the conceptual contexts can, in
return, trigger new qualitative contents. We can speculate that
only qualitative events are broadcast globally; global messages
may not be in the most universal code. If this were true, the
broadcasting arguments made before (2.xx) apply only to
qualitative events, and to the qualitative mental symbols of
non™qualitative concepts, beliefs, intentions and expectations.
This could explain many things: for example, the
extraordinary power of imagery in memory and emotion (Paivio,
1974; Horowitz, 1976; Singer, 1984); the great frequency of
visualizable prototypes of abstract categories (Rosch, 1975). The
power of prototypes in turn suggests the power of social
stereotypes: prejudice may consist of having standardized,
uncomplimentary mental images of the despised group, and a
failure to acknowledge the fact that reality more complex and
less imagible. In the realm of thinking, geometry has had great
impact in mathematics, even though since Descartes all
geometrical figures have been expressible algebraically. But
geometry can be visualized, while algebra cannot. All these
qualitative conscious experiences help to manipulate more
abstract non-qualitative entities.
6.6 Summary.
Starting from a consideration of William James' description
of the tip-of-the-tongue state, we noted that there are complex
representational states which dominate our central limited
capacity, and which act as goals. These goal contexts or
intentions are not conscious in the sense that they have no
conscious qualities (Natsoulas, 1982); they are experienced
differently from percepts, images and the like. One can view such
states as "goal contexts looking for conscious contents."
We have discussed some ways to test these claims empirically, and
explored their implications for spontaneous problem©solving, the
issue of significance, and the question of non©qualitative
conscious events.
In the following chapter we show how these ideas can lead
directly to a theory of voluntary control, a modern version of‹j ‹
William James' ideomotor theory.
Footnotes
1. Answers to the questions in 6.0:
1. pterosaurus,pterodactyl;
2. bionics, prosthetics;
3. loquacious, wordy, voluble, verbose, long-winded, etc.
2. Answers to the remote associate items from Mednick (1962):
1. sweet, 2. slow, 3. party, 4. ball,
5. memory, 6. salt, 7. lion.