Chapter Seven
Model 5:
Volition as ideomotor control of thought and action.
"We know what it is to get out of bed on a freezing morning
in a room without a fire, and how the very vital principle within
us protests against the ordeal. Probably most persons have lain
on certain mornings for an hour at a time unable to brace
themselves to the resolve. We think how late we shall be, how the
duties of the day will suffer; we say,"I îmustï get up, this is
ignominious," etc; but still the warm couch feels too delicious,
the cold outside too cruel, and resolution faints away and
postpones itself again and again just as it seemed on the verge
of bursting the resistance and passing over into the decisive
act. ...
Now how do we îeverï get up under such circumstances? If I may
generalize from my own experience, we more often than not get up
without any struggle at all. We suddenly find that we îhaveï got
up. A fortunate lapse of consciousness occurs; we forget both the
warmth and the cold; we fall into some revery connected with the
day's life, in the course of which the ideas flashes across us,
"Hollo, I must lie here no longer" --- an idea which at that
lucky instant awakens no contradictory or paralyzing suggestions,
and consequently produces immediately its appropriate motor
effects. ...
It was our acute consciousness of both the warmth and the
cold during the period of struggle, which paralyzed our activity
then and kept our idea of rising in the condition îwishï and not
îwillï. The moment these inhibitory ideas ceased, the original idea
exerted its effects.
This case seems to me to contain in miniature form the data
for an entire psychology of volition. ..."
--- James (1890, Vol. II, p. 524-5).
‹b ‹
7.0 Introduction.
7.1 Is there a problem of volition? Some contrasts between
similar voluntary and involuntary actions.
7.2 Voluntary action resembles spontaneous problem-solving.
7.21 Cooperating automatic systems control most of a
normal "voluntary" action.
7.22 We become conscious of underdetermined choice-points in
the flow of action.
7.3 The ideomotor theory in modern garb.
7.31 The fit with GW theory.
7.32 Voluntary action involves tacit editing of conscious
goals.
7.4 Evidence bearing on the ideomotor theory.
7.5 Explaining the voluntary-involuntary contrasts.
7.51 Slips: a losing horse-race between errors and
editing.
7.52 Counter-voluntary automaticity: the case of
"structural" slips.
7.53 Psychopathology: a repeated editing failure, perhaps
due to excessive control effort.
7.6 Wider implications.
7.61 What does it mean to make a decision?
7.62 Resistance,perceived effort and perceived control.
7.63 Ideomotor control of conceptual thought: A solution
to the puzzle of non-qualitative consciousness?
7.64 Fleeting goal images make accurate source attribution
difficult.
7.65 The relationship between a goal image and the action
it evokes may be highly variable.
7.7 Absorption and hypnosis as ideomotor events.
7.71 Absorption as a drop in competition for GW access.
7.72 Hypnosis as ideomotor control without competition.
7.8 Conflicts between goals.
7.81 A modern psychodynamics: Modeling competition
between opposing intentions.
7.82 Disavowed goals may be revealed in contradictions
between voluntary (edited) and involuntary
(unedited) expressions of the same goal.
7.83 The conflict-free sphere of conscious access and
control.
7.9 Summary.
7.91 Some testable predictions from Model 5.
7.92 Some questions Model 5 does not answer. ‹j ‹å
7.0 Introduction.
We begin our chapter on volition with the image of William
James on a cold winter morning, reluctantly trying to persuade
himself to get out of bed. For James, this image goes to the very
heart of the psychology of volition. He believed that a
successful act of will does not typically emerge from some
titanic inner struggle. Rather, he claims, we simply wait until
the conscious image of the action can emerge for some time
without competing images or intentions. At that moment the action
occurs automatically, spontaneously, and without struggle.
We will first consider whether there is a problem of
volition at all. To answer this question we seek contrasting
pairs of actions that differ only in that one action is voluntary
and the other, involuntary. These empirical contrasts can
constrain theory, just like the contrasts between conscious and
unconscious events (1.x). This evidence indicates that the issue
of volition is very real indeed. That is to say, the
voluntary/involuntary contrasts highlight core psychological
issues, such as automaticity due to practice, errors in speech
and action, and psychopathology. Further, we can borrow James'
solution to the problem of volitional control, and interpret it
easily in terms of global-workspace theory.
James explains conscious control of action by an îideo-motor
theoryï in which conscious goal images without effective
competition serve to organize and trigger automatically
controlled actions, which then run off without further conscious
involvement. For James, conscious contents are inherently
"impulsive"; everything else is automatic. The only conscious
components of action are:
(a) the "idea" or goal-image (really just an image of the
outcome of the action);
(b) perhaps some competing goal-image;
(c) the "fiat" (the "go signal");
and finally,
(d) sensory feedback from the action.
In this chapter we see how GW theory invites a natural
interpretation of James' ideomotor theory. One use of the GW
architecture is to have multiple unconscious systems inspect a
single conscious goal, and to compete against it if it is
inadequate. That is to say, the architecture allows multiple‹j ‹
unconcious criterion systems to îmonitorï and îeditï any conscious
goal or plan. This implies that any conscious goal image that is
conscious long enough to succeed in recruiting and executing an
action has been tacitly edited by multiple criteria, and indeed
we claim in this chapter that voluntary action îisï tacitly edited
action. Conversely, involuntary actions, like slips of the
tongue, are actions that îwould have beenï edited and changed, îifï
there had been enough time and capacity for unconscious editing
systems to be brought to bear upon the conscious action plans.
This conclusion has a wealth of implications for the
understanding of unintentional acts found in slips, automatisms,
and psychopathology. It even suggests a theory of hypnosis and
other "absorbed" states of mind, in which there is minimal
editing of conscious events.
Of course William James himself could not speak in these
terms, because of his resistance to the notion of unconscious
mental processes (1.x). But his ideas make perfectly good sense
in modern cognitive garb.
Before we proceed to develop these ideas, it is useful to be
clear about the issue of volition itself. Many behaviorists and
others have claimed that there is no problem of volition at all.
What evidence do we have to the contrary, that "the will"
matters?
7.1 Is there a problem of volition? Some contrasts between
similar voluntary and involuntary actions.
With the rise of physicalistic psychology at beginning of
this century, many psychologists tried to make the case that
there really is no question of volition, just as there was no
true scientific issue of consciousness (e.g. Watson, 1925;
Razran, 1965). Behaviorists and other physicalists at first
believed that any apparently "voluntary" action can be reduced to
a chain of conditioned, simple, îphysicalï reflexes. Later, when
reflexes proved too simple and rigid, the unit of behavior was
generalized to other stimulus-response relationships, but the
goal still remained to reduce voluntary, goal-directed actions to
simple, physical input-output relations (viz., Baars, 1986a).
This was thought to eliminate any scientific question of volition
once and for all. Was there any truth to this claim? Is there
indeed a scientific question of volition?
For an answer we can look to pairs of actions that appear
similar on the surface, but which differ in respect to volition.
That is, we can carry out a contrastive analysis on the issue of
volition, just as throughout this book we have contrasted
comparable conscious and unconscious events (1.xx). This is
helpful not just to answer the questions about volition raised
by Pavlov and Watson --- it also defines major constraints to be‹j ‹
satisfied by any theory of normal voluntary control. Any such
theory should be able to explain why, of two similar-seeming
actions, one seems to be voluntary and the other not.
There are some obvious examples of such contrastive pairs,
as we see in Table 7.1.
___________________________________________________
Table 7.1
__________________________________________________
Closely comparable contrasts involving volition.
Involuntary Voluntary
______________________________________________________
îNon©voluntary ïîaïîutomatisms: wanted, but not controllable in
detail.ï
Automatic components The same actions
of normal actions. before automaticity.
Reflexes. Purposeful imitations of
reflexes.
Actions controlled by The same actions initiated
brain stimulation by the patient.
of the motor cortex.
Autonomic functions. Autonomic functions under
(heart rate, peristalsis, temporary biofeedback
skin conductivity, etc.) control.
Spontaneous emotional "Social" expressions.
facial expressions
(Ekman, 1984)
îCounter©voluntary automatisms: unwanted. ï
Slips of speech and action. Purposeful imitations of slips.
Pathological symptoms: Purposeful imitations of
out©of©control actions, symptoms.
images, inner speech,
and feelings.
Regained voluntary control
after "practicing the symptom."
Voluntarily resisted habits Voluntarily controlled
(e.g. unwanted habits) automatisms.
____________________________________________________
îNon©voluntary vs. counter©voluntary events.
ïNotice first of all, that the "involuntary" events listed on
the left side of Table 7.1 are of two kinds. First, îautomatic
ïprocesses are part of every voluntary act, and while people
cannot control them in detail, they are perceived to be
consistent with our goals. We want them. A skilled typist does
not control each finger movement in detail; a skilled reader does
not perform letter identification consciously, etc. Yet because
automatisms serve our voluntary goals Table 7.1 calls these
"non™voluntary automatisms." On the other hand there are clearly
îcounter©voluntary actionsï such as slips of the tongue. Here, too,
there are automatisms at work, but they are perceived to be out
of control, unwanted, against one's will. These two kinds of
involuntary action may be closely related; for example, any
automatic component can become counter©voluntary simply by
resisting it. We can look at a word on this page, thereby
triggering automatic reading processes that are not under
detailed voluntary control. This wanted automatism can become
counter©voluntary simply by resisting it. Thus we can try to
resist the act of reading after looking at a word, or a knee©jerk
reflex after striking the patellar tendon. In this way any
automatism can be made counter©voluntary. This close
relationship between non©voluntary and counter©voluntary actions
makes it useful to consider both under the rubric of
"involuntary" activities. Whenever there is a possibility of
misunderstanding, we will choose an unambiguous term like
"automatic" vs. "counter©voluntary."
That being said, we can go on to discuss Table 7.1.
îSlips of speech and actionï
Imagine repeating a slip of the tongue you have just made.
The slip itself is experienced as involuntary; its imitation is
voluntary. And yet the two isolated actions are much the same as
far as an outside observer is concerned. Some famous slips by
A.W. Spooner illustrate the point:
(1) Instead of "our dear old Queen" --- "our queer old
Dean."
(2) Instead of the hymn, "Conquering Kings their titles
take." --- "Kinquering Congs their titles take."
(3) Upon dismissing a student, he intended to say "You have
deliberately wasted two terms, and you will leave by
the down train" --- but actually said, "You have â ‹j ‹
deliberately tasted two worms, and you will leave by â
the town drain."
Let us suppose Reverend Spooner actually made these slips
(there is some doubt: see Fromkin, 1980). Now imagine that
Spooner îrepeated each slipï immediately after making it, as
exactly as possible, so that it was said again by the same
speaker, in the same tone of voice, at the same speaking rate,
and so on. What is the difference between the slip and its
voluntary repetition? Surely there is no basic îphysicalï
difference, nor any real linguistic difference. The main
difference is psychological. In the first case, the utterance was
involuntary and unwanted; in the second, it was voluntary (Baars,
1985; in press).
But what a difference this invisible difference makes! In
the first case, the speaker fails to execute his intention. If he
becomes conscious of his error, he will experience îsurpriseï at
his own utterance. Now we can observe the whole panoply of
physiological reactions that make up the Orienting Response
(1.x). He may be embarassed and apologetic. Having failed to
carry out his intention, he may try again. If, like Spooner, he
is also head of one of the Cambridge colleges he may become a
figure of fun in student folklore. If he makes involuntary errors
so often that he can no longer function effectively, he may lose
his position, be examined for neurological problems, etc. None of
these consequences follow from doing physically identical
imitations of these slips, if they are voluntary. If Spooner were
voluntarily making the slip to amuse his audience, or if someone
quotes a slip in a discussion of voluntary control, none of these
consequences follow; nor is the speaker likely to be surprised by
the "slip."
Thus two identical actions may be psychologically quite
distinct, but not because of a difference in complexity, as the
early behaviorists thought. Voluntary actions are not just
complicated agglomerations of simple reflexes. Involuntary
components put together do not result in a voluntary act.
Something else is involved in volitional control. Consider two
more contrasts of this kind.
îThe loss of voluntary control with practice.ï
It is easy to see a voluntary act transformed into an
involuntary one: we only need to practice it to the point where
most of it fades from consciousness (5.xx). We have previously
pointed to experiments in which predictable skills are highly
overlearned, and which generally show a loss of voluntary control
(La Berge, 1984; Shiffrin & Schneider, 1977; Sternberg, 1963).‹j ‹å
All actions have involuntary components. Most details of
routine actions like reading or writing must be automatic: we
could never control their numerous details, given the limited
capacity of the conscious and voluntary system. Usually only the
novel features of an action are conscious and under voluntary
control (x.xx) (Reason, 1984). But non©voluntary automatisms can
sometimes become unwanted or counter©voluntary.
This becomes clear when we try to control "bad habits" that
have been practiced for years: almost everyone seems to have at
least one, whether it is over-eating, smoking, nervous gestures,
etc. These habits are characteristically difficult to control
voluntarily; they escape control especially when conscious
attention is directed elsewhere. No doubt unwanted habits have
multiple causes, but it is easy to demonstrate that sheer
automaticity makes it hard to stop an action once its normal
triggering conditions are given. As we pointed out above, looking
at a word îwithoutï reading it seems to be quite impossible (viz.,
Shiffrin & Schneider, 1977; La Berge, 1984). The very act of
looking at printed words seems to trigger automatisms; To block
them we must look away, degrade the word visually, or perhaps
focus on only a fraction of one letter. Sternberg's well-known
experiment in automatic memory search makes the same point
(Sternberg, 1963). The subject really cannot stop the search
process when the target letter is found; it just runs on to the
end of the memory set (see secton 1.xx). Reason (1983, 1984) has
presented a detailed analyses of catastrophic accidents showing
that many of them may be due to hard-to- control, highly
practiced automatisms that were triggered out of context at the
moment of the accident. Several of these accidents led to the
death of the person making the error --- about as strong an
argument for the involuntary nature of automatisms as we might
wish to have.
îInvoluntary automaticity involves a loss of conscious
access.ï
Loss of voluntary control over details of an action seems to
follow a loss of conscious access to the details. Langer and her
co©¨workers have conducted some elegant experiments to support
this point (e.g., Langer & Imber, 1979). These investigators
were pursuing the hypothesis that perceived competence affects
one's performance: the more skilled we think we are, the better
we perform --- providing that we cannot monitor our performance
directly. One way in which we lose touch with our own competence
is by automatization; when we become skilled readers, musicians,
or truck drivers, we lose conscious access to many details of our
own actions, and hence become more vulnerable to false‹j ‹
attributions about our own performance. This line of reasoning
led Langer and Imber (1979) to devise a simple coding task that
people could learn to the point of automaticity in a matter of
minutes. Letters of the alphabet were to be recoded into a two-symbol code;
the letters A©¨I were "a triangle plus the înïth letter
after A"; letters J¨©R would be "circle plus the înïth letter after
J," etc. Thus the letter "B" would be "triangle plus 2", "L"
would be "circle plus 3," etc. A preliminary group of subjects
reported that they were still conscious of task details after
recoding two sentences; after six sentences, they were no longer
conscious of the steps. The task had become automatic.
Langer and Imber (1979) now compared the effects of
conscious access and automaticity. A Moderate Practice group
recoded only two sentences, reported being conscious of details,
and was able to specify more steps in the task than the High
Practice group, which recoded six sentences and reported
automaticity. Now Langer and Imber devised an arbitrary task in
which some of the subjects would be called Bosses, others were
called Assistants, and a third group received no label. In fact,
the three groups did the identical task; the assumption was that
the labels would affect the self-confidence of the subjects.
Afterwards they were asked to do the coding task once again.
"Bosses" performed much as before, no different from the No Label
group. But "Assistants" now performed much worse ¨¨¨ îifï the coding
task was automatic. "Assistants" who were highly automatic in the
coding task made four times as many errors as before, and took
40% longer to finish. In the Moderate Practice condition, where
the coding task was not automatic and consciously accessible,
"Assistants" did as well as "Bosses."
The simplicity and effectiveness of this study is quite
remarkable. And the interpretation is quite clear: if we have no
conscious access to our own performance, and if some reliable
source of information îseemsï to indicate that we are doing quite
badly, we tend to accept misleading feedback because we cannot
check our own performance. With direct conscious access to our
performance we are much less influenced by misleading labels.
These results suggest that three things go together: losing
voluntary control over action details, losing consciousness of
them, and losing the ability to îmonitor and editï the details.
Indeed, the ability to monitor and edit a planned act may be the
essence of voluntary control (7.32).
While we may speak of "conscious" monitoring and editing,
the fact is, of course, that we are generally not conscious of
the rules and criteria by which we do our monitoring. If we find
a syntax error in inner speech, we do not consciously say, "Aha!
lack of number agreement between noun and verb!" Not even
linguists do that. Rather, we simply "know" immediately that the
conscious plan is in error. The rule systems that spot the error
are quite silent in their details. Thus it is not consciousness
that îdoesï the monitoring and editing; rather, conscious
experience of the event îfacilitatesï editing and monitoring by‹j ‹
unconscious rule systems, just as the GW architecture facilitates
the ability of many specialized processors to review a global
message.
Thus any complete theory of voluntary control must explain
the automaticity dimension: why, with practice, we lose both
conscious access to and voluntary control over the details of an
action.
îPathological loss of voluntary control.ï
Psychopathology is the study of repeated, dysfunctional
errors that are generally îknown to beï errors by the person who
makes them --- "slips" of action or experience that escape
attempts to control them, over and over again. Almost without
exception, psychopathology in the neurotic range involves a loss
of voluntary control over inner speech, feelings, mental images,
or overt actions. Loss of control over îinner speechï is a factor
in obsessive or delusional thinking, and in some auditory
hallucinations; out of control îbodily feelingsï play a role in
pathological anxiety, conversion hysteria, and depression;
uncontrolled îmental imagesï are at the core of phobias; and when
îactionsï run out of control we find compulsive or impulse-control
pathology.
We can illustrate all these points with a single patient who
suffered from a variety of symptoms. Consider Anna O., the
classical early patient of Breuer and Freud (18xx), who suffered
From a very severe case of conversion hysteria. As Erdelyi (198
describes the case (p. 20),
"Anna O. became Breuer's patient in 1880 at the age of 21
when, under the pressure of nursing her dying father, she
suffered a nervous collapse. She developed a veritable museum of
symptoms which included a labile (variable) pattern of
incapacitating paralyses of the limbs; depression and
listlessness; terrifying hallucinations of snakes, which
transmogrified into death's heads and skeletons; painful coughing
fits, especially in reaction to music; a period of severe
hydrophobia, during which she could not bring herself to drink
water; amnesias (blackouts) for recent events; a blinding squint;
severe paraphasia (loss of language ability); anorexia
(unwillingness to take food); and several other serious
dysfunctions."
It is the îloss of desired control ïthat makes these symptoms
pathological. Not moving one's limbs is quite all right if one
doesn't want to move them; depression and sadness due to a loss
is quite normal; strong squinting is a good idea in the middle of‹j ‹
a sun-drenched desert; even images of snakes and death's heads
can be quite normal for a reader of Gothic fiction (after all,
thousands of people voluntarily go to horror movies or read
Gothic tales); even amnesias for recent events can be normal when
we want to deliberately forget or ignore them. These events
become pathological when people do not want them. Those who
suffer from these symptoms try hard and often to master the
involuntary feelings, thoughts, actions, or images, but they fail
over and over again, in spite of often desperate efforts (e.g.
Horowitz, 1975 ab). It is not the îcontentï of the thoughts,
feelings, and actions that is problematic: it is their occurrence
out of an acceptable context, out of the perceived control of the
sufferer. Thus the issue of voluntary control is at the very core
of human psychopathology, and an understanding of psychopathology
must be grounded in an adequate theory of volition (see 7.8).
There is a clinical intervention that is sometimes very
effective, which seems to act directly on the mechanism of
voluntary control. This paradoxical technique is called "negative
practice," or "practicing the symptom" (e.g., Levine & Scheff,
1980). If a person has a specific phobia, he is told to
voluntarily bring forth the fearful images and thoughts; if he is
a stutterer, he is to try stuttering voluntarily whenever he
stutters spontaneously; and so on. Although this technique has
been known for decades, it has only recently begun to be
systematically tested in a variety of problems. Some of the
results are quite remarkable. Children who have stuttered for
years are told to stutter deliberately for 30 seconds each time
they do so involuntarily. As a result, they sometimes stop
stuttering in a day or two, with a 75% success rate (Levine and
Scheff, 1980; Levine, Ramirez, and Sandeen-Lee, 1982). There are
many cases in which the paradoxical technique works remarkably
quickly to stop anxiety attacks, compulsive actions, tics,
involuntary phobic images, La Tourette symptoms, and the like.
Here is a case where counter©voluntary automatisms are turned
into wanted but non©voluntary automatisms, just the opposite of
the case of "bad" habits discussed above. Of course, "practicing
the symptom" is not a cure-all. But it has been reliably observed
to stop pathological symptoms with remarkable speed, often after
years of helpless struggle.
Of theoretical interest here is the neat contrast between
voluntary and involuntary control in the paradoxical technique. A
habitual stutterer has typically struggled thousands of times
against the tendency to stutter. This repeated attempt to exert
voluntary control rarely works. The paradoxical intervention
requires him to stutter deliberately, îto do voluntarily what
normally happens involuntarilyï --- and rather magically, in many
cases the problem disappears. One fascinating possibility is that
the paradoxical intervention (which is, after all, only a switch
in the direction of voluntary effort) operates through the
voluntary control system. If this is true, then it may be that
the symptom itself is an error in voluntary control. Much
psychopathology may involve "errors of the will." These‹j ‹
speculations pose some important questions. We will return to
them when we attempt to model the voluntary-involuntary contrasts
of Table 7.1 (see 7.xx).
îVoluntary action is consistent with one's dominant
expectations.
ï
The cases discussed above --- slips, automaticity, and
psychopathology --- suggests that îcounter©voluntary action always
surprises the actor. ïThis is also true for non©voluntary
automatisms like reflexes when we resist them. Thus any
non™voluntary automatism is either surprising, or can be made to be
surprising when it is resisted. Under these circumstances the
automatism îïviolates dominant expectations (the dominant context
hierarchy) (viz., Baars, in press) º(Footnote 1). Conversely,
îvoluntary action ïseems always to be consistent with one's
dominant expectations.
There is direct evidence for this proposal from the study of
slips in speech and action (Baars, 1980 and in press; Reason,
1984). First, of course, we know that people often express
surprise when they make a slip. The Galvanic SKin Response (GSR)
is a well©established measure of surprise. GSRs monitored
immediately after experimentally elicited slips are quite large
when a sexually explicit slip is made, much smaller when a
neutral control slip occurs, and non-existent when the subject
makes a correct response (Motley, Camden & Baars, 1982). Thus
the more surprising (dominant context©violating) the slip, the
larger the GSR. Further, many slips are spontaneously
self-corrected immediately after they are made, again suggesting
that they surprise the speaker (Baars & Mattson, 1981). This
evidence supports the idea that îcounter©voluntary action
violates the expectations of the actorï, even when an outsider
might not notice anything unusual (Baars, in press, d).
îSome neurophysiological observationsï
Neuroscientists have never stopped using words like
"voluntary" and "involuntary" to describe some obvious phenomena.
Reflexes are obvious examples of involuntary actions; so are
autonomic functions like peristalsis, heart rate, sweat gland
activity, and the like, as opposed to the control of skeletal
muscles, which is voluntary in the usual sense. We now know that
autonomic functions can come under voluntary control at least
temporarily when people are given conscious feedback signals
activated by the autonomic functions (2.xx, 3.xx). Biofeedback‹j ‹
training seems to bring autonomic responses under the control of
the conscious/voluntary system. All these cases present obvious
contrasts between voluntary and involuntary control of the same
physical functions.
Another remarkable example of a neurophysiological contrast
between voluntary and involuntary control is cited by Penfield
and Roberts (19xx). These neurosurgeons used a low-voltage
electrode to explore the exposed cerebral cortex of conscious
patients, in order to identify and avoid critical areas where
surgery might cause serious damage. In one case as the surgeon
probed the motor cortex, the patient's hand moved, and the
patient was asked, "Are you moving your hand?" --- whereupon she
replied, with perfect accuracy, "No, doctor, îyouï are moving my
hand." How could the patient possibly tell the difference between
the brain mechanisms that were under "her own" versus the
surgeon's control? We do not know, of course, but her ability to
make this distinction suggests that there is a major difference
between voluntary and non-voluntary control.
In sum: Is voluntary control really a psychologically
significant issue? Facts like these indicate that it is indeed.
From here on we will assume that common sense is well justifi
in giving volition a fundamental psychological role (Footnote º2).
7.2 Voluntary action resembles spontaneous problem-solving.
In Chapter Six we worked out a way of understanding the
Conscious-Unconscious-Conscious (CUC) triad found in so many
types of problem solving. Thus, in answering a question we are
conscious of the question in detail, but not of searching for the
answer, though the answer is again conscious (6.xx). In creative
problem solving we are aware of the type of solution we need, but
not of the incubation process which eventually brings it to
awareness. And so on. Further, we have addressed the whole
question of what is meant by an îintentionï by considering the tip-
of-the-tongue (TOT) state, concluding that even as we are
searching for the right word, there is a state of mind which
constrains the search, which constrains limited capacity, but
which does not have qualitative conscious contents like color,
texture, or flavor. This "intention to say so-and-so" was called
a dominant goal context (6.xx).
Voluntary control resembles spontaneous problem-solving in
many ways. As James suggests, in voluntary action a conscious
goal image may be carried out unconsciously, and the results of
the action often become conscious again (7.xx). For illustration,
we will ask the reader to turn this book upside-down. (It is
helpful to actually carry out this little experiment in‹j ‹
self-observation.) Clearly the reader is conscious of the request
to turn the book upside-down, and perhaps of some visual image of
how this might be done. However, the request is ambiguous: is the
book to be turned in the horizontal or the vertical plane? This
ambiguity may be conscious for some readers and unconscious for
others. The mechanics of controlling hand and arm muscles are
surely not conscious, although îchoice-points and obstaclesï (how
do I turn the book upside-down without spilling my coffee?) may
be conscious. And of course the results of the action will be
conscious.
Further, there is a set of îconstraintsï on the action,
represented in GW theory by the îdominant goal contextï, which are
not likely to be conscious at any time during the action (Figure
6.x). We probably turn the book over with maximum economy of
movement, rather than sweeping through the air with graceful,
elaborate gestures. Then there are constraints imposed by the
need to maintain physical control of the book; we are not likely
merely to flip it up into the air and let it fall helter-skelter.
Even further, there are constraints of convenience, such as
keeping track of one's place even while indulging in this little
thought experiment. We must stop reading while the book is being
moved, and we make automatic postural adjustments to balance the
changing forces on the body. Finally, there may be social
considerations --- if we are in public, is anyone watching our
peculiar behavior? While some of these considerations may be
momentarily conscious, many of them will be unconscious, but they
still serve to constrain the action.
In a real sense the action that results from this complex
set of momentary conscious and unconscious constraints is a
îsolutionï posed by îproblemsï triggered by the conscious goal, and
bounded by numerous physical, kinetic, social, and other
contextual considerations. It makes sense therefore to treat
voluntary control as a kind of problem-solving (6.0).
7.21 Cooperating automatic systems control most of a
normal "voluntary" action.
The bulk of spontaneous problem-solving is unconscious
(6.xx). The same is surely true of voluntary actions. Much of our
intention to perform a particular act must be formulated
unconsciously, and the muscular effectors and subgoals needed to
carry out the intention are also largely unconscious. Thus many
systems cooperate in creating a voluntary act. It is good to
keep this great amount of cooperative processing in mind during
the coming discussion, which will focus mostly on the îcompetitiveï
aspects of voluntary control.
Notice, by the way, that the same systems may cooperate most‹j ‹
of the time, only to begin competing when the action runs into
trouble. If many systems work together to structure normal
speech, a slip of the tongue will seem erroneous to some but not
all of those systems. When Spooner slipped into "our queer old
Dean," he made no error at all lexically, phonetically,
syntactically, or even in pronunciation rules. The only systems
able to detect the errors are semantic and pragmatic: that is,
the systems that control meaning and communicative purpose. Those
are the only levels violated by the slip. It would seem to follow
that those systems begin to compete against the error, while the
others continue to cooperate.
7.22 We become conscious of underdetermined choice©points in
the flow of action.
If we are unconscious of these routine, cooperating systems,
what are we conscious of? Our previous discussion (5.xx) suggests
that the most informative aspects of action should be conscious:
that is, those that are unpredictable and significant. It is the
îunderdetermined choice©points ïin the flow of action that should
be conscious most often. In speech, hesitation pauses are known
to occur at points of high uncertainty (Goldman©Eisler, 1972).
Clearly, making people conscious of their routine speech will
slow down or interrupt the flow of speech, because previously
parallel automatisms are now channeled through the
limited™capacity bottle©neck; thus hesitation pauses may reflect high
conscious involvement. There is considerable independent
evidence for limited©capacity©loading events at junctures in the
flow of speech, such as clause and sentence boundaries (Abrams &
Bever, 1969). These junctures are likely to be points of high
uncertainty. While this evidence does not prove conclusively that
there is more conscious involvement at these points, it makes the
hypothesis plausible.
Given these considerations, we can now explore the ideomotor
approach to voluntary control.
7.3 The ideomotor theory in modern garb.
James' ideomotor theory fits neatly into the global-
workspace framework. According to this view, a single conscious
goal-image, if it does not meet with competition, may suffice to
set off a complex, highly coordinated, largely unconscious
action. For William James the ideomotor concept emerged from a
puzzle in the experience of starting an action: Do we ever
experience any command at all? Introspective reports on action
commands were vague and contradictory, and this question became a‹j ‹
major source of controversy between Wundt, James, and the
Wuô"rzburg School (James, 1890, pp.). James suggested that there
is, in fact, no experience at all of commanding an action;
rather, an action is organized and initiated unconsciously,
whenever a certain goal image becomes conscious without effective
competition.
We can partition the ideomotor theory into five interacting
hypotheses:
1. The îConscious Goal Imageï Hypothesis is the one just
stated, that all actions are initiated by relatively simple and
momentary images of the goal. For many actions these images may
be visual, because the visual system is very good in representing
spatial properties of action. However, auditory, tactile, taste,
or smell images are not ruled out. The act of walking to the
kitchen to prepare lunch may be initiated by a taste and smell
image of an attractive peanut-butter-and-jelly sandwich.
2. The îCompeting Element Hypothesisï is the notion that
competing events may drive the goal image from consciousness.
Competing events include conscious goal images as well as the
non-qualitative intentions we have discussed previously (6.0).
This idea has many important implications. It allows new
conscious thoughts or images to interfere with the planning of an
action, and it also permits editing of the goal by many different
intentional goal systems.
3. The îExecutive Ignoranceï Hypothesis suggests that most
detailed information processing is unconscious and that executive
processes have no routine access to the details of effector
control (Greene, 1971; Baars, 1980). Control of the muscles that
are used to carry out an action is simply unconscious.
4. The îAction Fiat Hypothesisï claims that the moment of
willingness to execute the action may be conscious, especially
when the time to execute is non©routine. (James calls this the
"fiat," the mental permission to start the act).
5. Finally, the îDefault Execution Hypothesisï is the tendency
of the goal image to execute in the absence of any effective
competition --- "by default." This is really just another side of
the Competing Elements Hypothesis, but it is useful to emphasize
it with a special label.
In addition to these five points, we should be reminded that
îsubïgoals needed to accomplish the goal may become conscious if
the goal cannot execute automatically (7.31). But let us suppose
for now that all subgoals are automatic and unobstructed, so that
they can execute without further conscious involvement.
To make these abstractions easier to imagine, take the
example of retrieving a word, intending to say it, and then
saying it. We have previously noted that complex activities like‹j ‹
word retrieval and speaking involve many separable components.
Because of the limited capacity of consciousness we cannot afford
to think consciously about many details in the act of speaking;
we want to access all components of speaking at once, so that
"the speech system" behaves as a single processor. But when we
change from speaking to listening, or from speaking to eating, we
may want to îdecomposeï the unitary speech system, to reorganize
its components into new configurations for listening, chewing
food, inner speech, and the like.
The ideomotor theory suggests that the "speech processor" as
a whole must be recruited, organized, and triggered by a single
conscious goal image. This image is itself controlled by a
higher-level goal structure --- for example, the reader's general
willingness to go along with demonstrations in this book. The
following example explores ideomotor control in detail.
î1. Conscious Goal Images can activate unconscious goal
structures.ï
If we ask the reader: "What are two names for the winged
dynosaurs that lived millions of years ago?" The question is
obviously conscious. Now, according to the ideomotor theory, this
conscious experience initiates an intention to retrieve a word
that matches the intention. Further, the conscious question
triggers unconscious search processes which produce candidate
words that may match or mismatch the intention (6.x). Because the
words are rare, the momentary intention is likely to be prolonged
into a tip-of- the-tongue state.
In GW theory, a "conscious goal image" is of course a
global, consistent representation that provides information to
numerous specialized processors (2.2). It is not surprising that
a conscious goal would trigger local processors that control the
muscles that carry out the goal. Indeed, we have argued early on
(1.xx) that specialized processors are often goal-addressible:
they are activated by goals. One nice feature of the GW system
is that the goal image can be quite arbitrary or fragmentary,
since it is the specialized processors themselves that have the
real "intelligence" of the system, and which interpret the
implications of the goal image in their own ways. Note that the
goal image can trigger both the subordinate specialists able to
carry out the action îandï the intentional goal context which
constrains planning and execution without itself becoming
conscious (Figure 7.3).
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
Insert Figure 7.3 about here.
©©©©©©©©©©©©©©©©©©©©©©©©©©©©©
In fact, the goal image itself results from yet a higher-
level goal context. Speaking is normally in the service of some‹j ‹
other goal --- communicating a thought, calling attention to
oneself, gaining information --- which is, in its turn, in the
service of even higher-level goals.
î2. Conscious Goal Images can also recruit a coherent set of
action schemata and effectors, even though we have Executive
Ignorance about the details of those processors.ï.
Now let us suppose that the reader has recalled the name
"pterosaurus" (or "pterodactyl") as an answer to the question
posed above. This is a conscious representation of the word. Now,
how do we recruit the largely unconscious systems that control
pronunciation of this difficult word? It is useful to recall here
how complex and fast-moving the speech apparatus really is, and
how little of it is accessible to awareness at any single time
(Executive Ignorance). It seems plausible that the conscious
word, in combination with a goal context, can recruit and
organize the complex effector system needed to pronounce it.
îExecutive Ignorance of action detailsï is implicit in GW
theory. As long as the details of action are unconscious, GW
theory suggests that executive goal systems operating through the
global workspace do not have direct access to such details.
î3. Default Execution: Given a compatible Dominant Goal
Context, a conscious goal tends to execute automatically.ï
Once "pterosaurus" becomes conscious in the presence of an
intention to say the matching word, something rather magical
happens: we suddenly notice that our mouth has begun to pronounce
the conscious word. The intervening steps of motor control are
simply not conscious. In James' words, "consciousness is
impulsive" --- unless, of course, other goal systems begin to
compete for access to consciousness.
The notion that specialized processors tend to execute
automatically, in the absence of contrary conscious messages, is
already implicit in basic GW theory. There is nothing to stop an
unconscious processor from executing an action except for
contrary conscious images and intentions. If those are absent, we
can expect actions to run off by themselves.
î4. The Competing Element Hypothesis: Conscious contents can
be edited by multiple unconscious goal systems.ï
Suppose the reader first retrieves "tyrannosaurus", instead
of "pterosaurus"? Clearly we do not want to execute this
incorrect goal image. Various knowledge sources should interfere
with its execution: some may remind us that "tyrannosaurus" is‹j ‹
too long, or that it has a different meaning. Such
contradictory knowledge should have access to the global
workspace, to compete against the incorrect conscious goal image.
GW theory thus suggests that editing of flawed conscious plans is
not some "added-on" capacity, but an integral aspect of the
architecture of the cognitive system.
In GW terms, the goal image may also set off processors that
generate competing goal images. Perhaps some of these contradict
the first goal image, or present alternatives to it (see Figure
xx). If some unconscious system detects a bad error in the goal
image, it may trigger competing images that act to destroy the
flawed conscious goal --- to edit and correct it. But once a
single goal image wins out long enough, the dominant goal image
will be executed. Its details are obviously off the global
workspace, and hence unconscious. Figure 7.x presents this series
of events in detail.
GW architecture supports editing of a global plan by
potentially îanyï rule system. Take a single sentence, spoken by a
normal speaker. Errors at any level of control can be detected îifï
the sentence becomes conscious (e.g. MacKay, 1980). There are so
many ways errors can creep into a sentence, and a correspondingly
large number of unconscious rule systems that constrain
successful sentences. There are many ways to be wrong, and only a
few ways to be right by all criteria. Thus we can very quickly
detect errors or anomalies in pronunciation, voice-quality,
perceived location of the voice, acoustics, vocabulary, syllable
stress, intonation, phonology, morphology, syntax, semantics,
stylistics, discourse relations, conversational norms,
communicative effectiveness, or pragmatic intentions of the
speaker. Each of these aspects corresponds to very complex and
highly developed rule-systems, which we as skilled speakers of
the language have developed to a high level of proficiency (e.g.
Clark & Clark, 1977). Yet as long as we are conscious of the
spoken sentence we bring all these rule-systems to bear on the
sentence --- we can automatically detect violations of any of
them, implying that the sentence is somehow available to all of
them (2.xx).
In principle, the set of "editing systems" is an open set.
We can always add some new criteria for correct performance. This
is one reason to suggest that conscious goals are "universally
edited." Obviously the most effective competition is from goal
contexts in the Dominant Goal Hierarchy, since these already have
GW access during preparation and execution of the action (4.xx;
6.xx; Figure xx). But entirely novel aspects of the action can in
principle be monitored and edited by onlooking processors,
providing they can compete for access to the global workspace.
Thus if one prepares to say a sentence, and suddenly a buzzing
fly darts into one's mouth, the action can be aborted even though
this situation was not anticipated as part of the goal context.
Novel considerations can compete against the global goal.
‹j ‹å If all conscious goal images are inherently edited by
onlooking processors, it follows that conscious goals that are
actually carried out îmust have been tacitly editedï by relevant
systems. Further, because îanyï system can potentially compete
against the goal image, we can talk about this system as allowing
îuniversalï editing. In section 7.32 we argue that this is indeed
the criterial property of voluntary action: Voluntary action is
action whose conscious components have been tacitly edited prior
to execution.
î5. The Action Fiat Hypothesis: The moment of execution ââmay
be under conscious and voluntary control.ï
We wait to say "pterosaurus" until we get a conscious
signal; by contrast, in speaking a stream of words, we rarely
seem to control the onset of each individual word consciously.
But with an isolated word or action, given enough lead time, we
can report fairly accurately our intention to execute the action
at some specific moment. One key difference is whether the moment
of onset of the action is automatically predictable; if it is, it
is rarely conscious; but if the moment of onset is unpredictable,
conscious control becomes more likely.
How should we represent the Action Fiat Hypothesis in GW
theory? If goal images tend to execute automatically, it makes
sense to suppose that timing an action involves inhibiting
execution up to the right moment, and then releasing inhibition.
Presumably, specialized processors sensitive to timing act to
hold up execution of a goal image until the right moment (Figure
x).
î6. Mismatch, surprise, and corrective feedback.ï
Conscious feedback resulting from an action can reveal
success or failure to many unconscious goal systems, which may
then develop corrective measures.
Imagine trying to say "pterosaurus" and actually saying,
"ptero... ptero ... pterosaurus" --- a momentary stutter that is
quite common in normal speech. Although we have no routine
conscious access to the complex articulators and timing systems
that control speech, it seems that those specialized systems îdoï
have access to conscious events. In general, when we allow errors
to become conscious, chances are that we can learn to avoid them
in the future. In GW theory, consciousness of feedback from the
flawed action sets into motion unconscious specialists that
attempt to repair the dysfluency.
When we notice a speech error consciously, we often "repair"‹j ‹
it quickly (Clark & Clark, 1977; MacKay, 1980), but we are rarely
conscious of details of the repair. Responding to overt errors is
similar to anticipatory editing of covert errors, except that
editing takes place before the action is executed (7.32).
Correction of overt errors is useful in preparing for a more
error©free performance next time around.
We have previously suggested that surprising events may
involve disruptions of one level of context, even while higher
levels are undisturbed (x.xx). Thus repair of contextual
violations may start at a higher level than the level that was
violated. The same thing may be true of errors in action. If we
stutter, the error is at the level of articulation, but higher
levels of control --- phonemic, lexical, syntactic, etc. --- are
unaffected. Thus higher-level goal systems may seek another way
to reach their goals. It is rarely the case that the entire
dominant goal hierarchy is disrupted, fortunately for us (9.xx).
In sum, the Jamesian ideomotor theory can be incorporated
straightforwardly in GW theory. In fact, it is difficult to see
how one could believe that a conscious goal image is executed
unconsciously without the concept of a distributed system of
intelligent processors, able to interpret and carry out the
relatively crude conscious goal.
The tip-of-the-tongue experience for "pterosaurus" helped to
illustrate the intuitive plausibility of the ideomotor theory,
and the rather nice fit with GW theory. But it does not provide
proof. In section 7.4 below we will discuss the evidence for or
against each hypothesis, and its implications for a broad theory
of voluntary control. But first, we are ready now to make a basic
theoretical claim about the nature of voluntary action.
‹F ‹
7.32 Voluntary action involves tacit editing of conscious goals.
If there is indeed universal editing of conscious goals, the
conscious aspects of any action îmust have beenï tacitly edited for
consistency with one's goal hierarchy before the action was
performed. Take the example of premeditated murder. If a normal,
rational person has thought for weeks about committing murder,
and proceeds to do so, we immediately make the inference that
contrary thoughts îmust have beenï entertained and rejected: that
the murderer must have anticipated the chances of being caught,
the likely disapproval of others, and perhaps the suffering of
the victim and his family. That is, we immediately infer that
competing alternatives will have been evaluated for any conscious
impulse to action that was considered for some time, especially
if the action has heavy potential costs. If the action was taken
in spite of these "editing" thoughts, we make inferences about
the value system of the murderer, or about mitigating
circumstances. The important point for us here is the idea that
conscious impulses are presumed to have been edited before
action, assuming there was enough time to do so.
What components of action are likely to be conscious, and
therefore tacitly edited? The theoretical answer fits our
previous supposition (7.22): we tend to be conscious of those
aspects of action planning that are novel, informative,
significant, or conflictful (see chapters 4.0, 5.0, and 6.0).
Those features that require the integrative capacities of a
global workspace system are precisely those that are likely to be
conscious ©©© and those are of course exactly the ones that are
likely to need editing.
A major claim in this chapter is that îvoluntary action is in
its essence, action whose conscious components have been edited
before being carried outï. In contrast, counter©voluntary actions
such as slips are editing failures: actions that îwould have been
edited and changedï had there been an opportunity to do so before
execution. Of course, most components of a normal action are
unconscious; these components cannot be globally edited before
execution. However, even automatic components of action must
have been conscious at some time in the past. Therefore they must
have been implicitly edited at that time to make them consistent
with the Dominant Goal Hierarchy. Of course if automatic
components run into trouble, they tend to become conscious, and
can be edited.
There is direct evidence for an editing capability of this
kind, and when we turn to the voluntary-involuntary contrasts
(7.xx) we will find that the major difference between closely
matched voluntary and counter©voluntary actions is this ability
to edit.
‹j ‹å The five main parts of the ideomotor hypothesis seem to fit
the GW framework remarkably well. The interpretation does not
seem forced or awkward. Further, as we look at the world from the
resulting point of view, many other pieces of the puzzle begin to
fall into place (7.x). The payoffs of bringing the ideomotor
concept into our model are therefore attractive, and the
theoretical costs seem minimal. But what about empirical support?
Do the facts justify our taking the ideomotor theory seriously?
Let us see.
7.4 Evidence bearing on the ideomotor theory.
A good deal of evidence is consistent with the ideomotor
theory, though the case is not air©tight. Consider the following
points:
7.41 Evidence for the impulsivity of conscious goal images.
The "Chevreul pendulum," a classic demonstration of the
impulsive force of conscious goals, has been used has been used
since the 19th century to persuade hypnotic subjects of the power
of their own unaided thoughts (James, 1890). One simply takes a
pendulum consisting of a string with a weighted bob at the end,
and tries to hold it completely steady. Now, while trying to keep
it steady, the subject begins to îthinkï of the pendulum as
swinging away from and toward him, on a North-South axis. Without
any perceived effort, the pendulum will begin to swing North and
South. Again, making every effort not to move the pendulum, the
subject begins to imagine it swinging right to left, in an
East-West direction. The pendulum soon begins to follow the
subject's thoughts, even though there is no noticeable effort or
movement of the hand! It seems as if conscious images are more
powerful than deliberate intentions.
It is not easy to adapt this classical demonstration to the
rigors of modern investigation. The ideomotor theory needs a
great deal more empirical support than is provided by
demonstrations such as this. But it is difficult to doubt that
there are conscious events related to goals: people can report
their own conscious thoughts and images regarding a planned
action, and usually predict their actions accurately in the very
short term. But do those conscious events actually trigger off
actions? This is difficult to be sure about, especially in view
of the fact that some fleeting conscious goals that are difficult
to report may evoke action (1.xx).
We do know that there is a momentary increase in mental
workload immediately before the onset of an action (Keele,
1973). This is consistent with the idea that there is at least a
momentary conscious goal prior to action. Libet (1985) has‹j ‹
presented arguments that we may become conscious of an action
only îafterï the brain events that immediately trigger it. But this
cannot be true in every case: surely there are many cases where
people are conscious of what they are about to do seconds or
hours before they do it, as shown by the fact that they can
accurately predict their actions beforehand. The reader may make
a conscious and reportable decision right now to turn the page,
and actually do so: This is hardly surprising, but any theory
that cannot handle this elementary fact is incomplete.
More evidence for the influence of conscious goals comes
From the experimental literature on mental practice, showing th
consciously imagining an action can improve performance as much
as actual physical practice (Maltzman, 19xx). Conscious imaging
of goals is used extensively in clinical practice and to improve
athletic performance (Singer, 1984). There is no doubt that
conscious images of goals can have powerful influence on
effective action.
Further, we know that the opposite case also holds: îlossï of
conscious access to an action can lead to a loss of control.
Langer and Imber (x.xx, 1979) showed that automatization of a
coding task leads to a loss in ability to evaluate one's own
performance, and Reason's analysis of errors and accidents also
shows a plausible relationship between automaticity and loss of
control (x.xx, Reason, 1984). Automatization presumably means
that goal images become less and less available, and therefore
the actions themselves become less and less modifiable (x.xx).
Some of the most direct evidence for the role of conscious
events in influencing action comes from conscious priming of
experimentally evoked slips of speech and action. There are now
several techniques for eliciting these slips in the laboratory
(e.g. Baars, 1980, 1985, in press). One of these techniques uses
phonological priming --- that is, conscious exposure to words
that resemble the slip --- to elicit spoonerisms. Here is an
example. The reader can ask someone to repeat the word "poke"
about half a dozen times, and then ask, "What do you call the
white of an egg?" Most people will answer, "the yolk" even when
they know better. They have evidently been primed by the
conscious word "poke" to retrieve a similar-sounding word from
memory (Kimble & Perlmuter, 19xx). This technique may work
because it duplicates the normal effect of conscious goal images,
which prime the action to be taken.
In general, spoonerisms can be elicited by consciously
priming the speaker with word-pairs that resemble the predicted
error (Baars, 1980a, in press). Thus the slip îbarn doorï - îdarn
boreï can be elicited by showing a subject a series of word-pairs
like îdart boardï, îdark bowlï, îdot boneï, etc. Because subjects do
not know ahead of time which word-pair they must say out loud,
they must be prepared to say each one. This state of readiness
apparently primes the system to make an error when the phoneme
pattern is switched. ‹j ‹å
There are several other techniques for eliciting errors. All
of them seem to create competing speech plans, compelling
subjects to choose very quickly between the two alternatives
(Baars, 1980b). Sentence errors like the following are triggered
by creating uncertainty about the order of two phrases in a
target sentence. If people are unsure about whether to say, îShe
touched her nose and picked a flowerï, or îShe picked a flower and
touched her noseï, they are likely to say inadvertently, îShe
picked her nose ...ï. There are several ways to create this
uncertainty. The easiest is to present the stimulus sentences,
and after each one simply signal the subject either to repeat the
previous sentence in the order given, or to reverse the phrases
of the sentence. This technique produces predictable word™exchange
slips at an adequate rate. Materials can be designed so
as to elicit almost any involuntary statement from the subjects
(Baars, 1980a; in press).
All slip techniques to date create a state of readiness in
the speech system to act in a certain way ©©© they create goal
contexts. Once this is done, we can ask whether adding a
conscious image related to the target slip will increase the
chances of the slip. For example, if we gave people the conscious
word pair "terrible error," would that increase the chances of
the slip "bad goof"? Motley & Baars (1979 a) showed that it does
indeed. Further, if people are presented with a social situation
such as the presence of an attractive member of the opposite sex,
slips related to the situation are made much more often (see
Baars, in press). In all these cases, a conscious prime coming
just before a potential related slip will sharply increase the
chances of making the slip. This suggests that conscious events
can help recruit actions. While this evidence does not totally
confirm the impulsive force of conscious goal images, it does
support this part of the ideomotor theory.
7.42 Evidence for editing by global competition.
If a momentary conscious goal image is necessary to set up
and trigger an action, competing conscious events should be able
to delay or inhibit it. Everyday experience fits this pattern
well. If we ask someone to remember a difficult word, and then
interrupt with any other demanding conscious task, the desired
word will simply not come to mind long enough to allow the person
to say it. This is obvious and cannot be ignored. Thus editing
may simply take place by competition for access to the global
workspace, coming from processors that can detect the erroneous
goal image. This competition can then keep the error from
dominating the global workspace long enough to recruit and‹j ‹
trigger action. It is theoretically pleasing that we need add no
new elements for editing to take place: it is simply another
application of the general fact that the GW architecture permits
local specialists to compete against global messages.
Other observations are consistent with this view. Thus
Meichenbaum and Goodman (1971) have shown that impulsive
children can use inner speech to improve self-control. If
impulsivity consists of having very powerful conscious goal
images that do not encounter immediate competition, then training
children to use conscious inner speech may help them to compete
against the undesirable goal image. The impulsive goal images may
become less consciously available, and have less time to organize
and execute unwanted actions. On the other side of the editing
coin, Langer & Imber's findings (discussed above) indicate that
practicing a task to the point of automaticity leads to a loss
of ability to monitor the action. Apparently conscious goal
images are less and less easy to monitor as an action becomes
more and more automatic (Pani, 1982; see 1.xx).
Another source of evidence for anticipatory editing comes
From experimentally elicited slips. One can get subjects to ma
slips of the tongue that violate the general rules of language or
usage; these slips can then be compared to very similar slips
that do fit the rules. Thus, in the laboratory people will make
slips like:
(1) îdarn boreï --- îbarn doorï (meaningful words)
(*) (2) îdart boardï --- îbart doardï (nonsense)
(3) înery viceï --- îvery niceï (syntactically correct)
(*) (4) îvice neryï --- înice veryï (wrong syntax)
(5) îlice negsï --- înice legsï (sexual comment that may be
socially inappropriate)
(6) îreel fejektedï --- îfeel rejectedï (depressed comment)
Likewise, we can elicit word-exchange slips like:
(*) (7) îShe touched her nose and picked a flower.ï
--- îShe picked her nose ...ï (socially embarassing)
(*) (8) îShe hit the ball and saw her husband.ï
--- îShe hit her husband ...ï (aggressive affect)
(*) (9) îThe teacher told the myths and dismissed the stories.ï
--- îThe teacher dismissed the myths...ï (hard to pronounce).
(*) (10) îShe looked at the boy and talked softly.ï
--- îShe talked at the boy and looked softly.ï
(semantically anomalous).
(*) (11) îIs the gray sea below the blue sky?ï
--- îNo, the blue sky is below the gray sea.ï (false)
By designing slips that violate some level of control, and
comparing them to very similar rule-îgovernedï slips, we have found
a number of cases where the rate of rule-violating slips drops
precipitously, sometimes even to zero (e.g., Baars, 1980a; Baars,
Motley and MacKay, 1975; Baars & Mattson, 1982; Motley, Camden &
Baars, 1979). All starred (*) slips listed above violate generic
rules, and these slips show lower rates than similar slips that
obey the rules. If the drop in rule-violating error rates is due
to some editing process, the fact that this occurs with so many
different rule-systems --- pronunciation, phonology, lexical,
syntactic, social, etc. --- supports the idea of îuniversalï
editing.
Ordinarily we think of "editing" as a review process in
which someone like a newspaper editor checks the output of a
journalist against certain criteria --- criteria like linguistic
adequacy, fit with editorial policy, and the like. In general
"editing" seems to involve two separate entities, one of which is
able to detect errors in the output of the other system.
To show that editing in that sense occurs in normal speech
production, we need to demonstrate that people in the act of
speaking can detect mismatches between a speech plan and their
criteria. Motley, Camden & Baars (1982) report that for a task
eliciting sexually expressive slips (îlake muvï - îmake luvï, îbice
noddyï - înice bodyï), there is a large and rapid rise in the
electrical skin conductivity on sexual slip trials îeven if the
slip is not actually madeï. On neutral control items there is no
such effect. Since the Electro-Dermal Response is one of the
standard measures of the Orienting Response --- a reliable
physiological index of surprise --- these results suggest that a
mismatch was detected even when the slip was successfully
avoided. Thus egregious errors can be detected even before they
are made overtly, and suppressed. This is exactly the notion of
editing suggested above.
We cannot be sure in these experiments that the edited
speech plan was conscious, but we do know that conscious speech
errors can be detected by many largely unconscious criteria. Not
all errors in spontaneous speech are detected, not even all overt
errors (MacKay, 1980). But once speakers become conscious of an‹j ‹
error they are likely to correct it. In fact, normal speech is
marked by great numbers of overt self-corrections or "repairs"
(Clark & Clark, 1977). In any case, only part of the process of
error-detection and correction is conscious and reportable.
Certainly the slip itself is so, often, but detailed mechanisms
of detection and correction are not. Therefore, even though we do
not know for sure that the edited slips in the above experiments
were conscious, we can certainly suggest that unconscious editing
of conscious errors occurs quite commonly.
7.43 Evidence for Executive Ignorance.
Try wiggling a finger: where are the muscles located that
control the finger? Most people believe that they are located in
the hand, but in fact they are in the forearm, as one can tell
simply by feeling the forearm while moving the fingers. What is
the difference between pronouncing /ba/ and /pa/? Most people
simply don't know. In fact, the difference is a minute lag
between the opening of the lips and the beginning of vocal cord
vibration. These examples can be multiplied indefinitely. We
simply have no conscious, reportable access to the details of
action.
7.44 Evidence for the Action Fiat.
We can prepare for an action and suspend execution until
some "go" signal. The time of the "go" signal can be conscious,
witness the fact that people can tell us when they will execute
the action. In that sense, people clearly have conscious access
to, and control of, the "action fiat".
The separation between îpreparationï and îexecutionï seems to
exist even when execution is not delayed. All actions seem to
have these two phases. For example, in the cat, where the
neurophysiology of action control has been worked out to a
considerable extent, there seems to be a natural division between
preparation and execution. As Greene (1972) writes:
"When a cat turns its head to look at a mouse, the
angles of tilting of its head and flexion and torsion of its neck
will tune spinal motor centers in such a way that its brain has
only to command 'Jump!' and the jump will be in the right
direction. ... the tilt and neck flexion combine additively to
determine the degrees of extension of the fore and hind limbs‹j ‹
appropriate to each act of climbing up or down, jumping onto a
platform, standing on an incline, or peering into a mousehole;
the neck torsion regulates the relative extensions of left and
right legs when preparing to jump to the side. These postures
must be set as the act begins; for if they were entirely
dependent upon corrective feedback, the cat would have stumbled
or missed the platform before the feedback could work. A few of
these reflex patterns of feedforward are adequate for the
approximate regulation of all feline postures and movements
required in normal environments for a cat..."
When is the action fiat conscious? We can suggest that this
depends on predictability of the time of action, just as
consciousness or automaticity in general depends upon the
predictability of any action subsystem. The action fiat should be
conscious when the time of execution is non-routine.
7.45 Evidence for Default Execution.
How do we know that conscious goals tend to be executed in
the absence of contrary conscious or intentional events? Part of
the reason comes from the kind of demonstration of automaticity
we suggested before: try looking at a word without reading it, or
in the case of rapid memory scanning, try stopping automatic
memory search before the end of the list. (Sternberg, 1963; x.xx)
Or consider once again the absent-minded errors collected by
Reason and his colleagues (5.xx). Reason reports that îstrong
habit intrusionsï occur in the course of normal actions when the
actor is absent©minded or distracted, hence unable to pay
attention (to be conscious of the relevant aspect of action).
These cases argue for Default Execution. It seems as if a
prepared action executes even when it should not, if contrary
conscious events do not block the faulty action. This failure to
block a faulty goal image can have catastropic consequences.
Reason (1983) has analyzed a number of accidents like airplane
crashes and road accidents, and concludes that many of these
disasters may be caused by the intrusion of automatic processes,
in combination with a low level of conscious monitoring.
A child of six knows how to keep such errors from happening:
you have to îpay attentionï to what you're doing. That is, you must
be conscious of the novel circumstances and goals. When we pay
attention, erroneous Default Executions do not occur. However, it
seems likely that the same principle of Default Execution is used
to execute îcorrectï actions most of the time. We seem to
automatically carry out conscious goals, unless contrary images
and intentions block the conscious goals.
‹j ‹å
7.5 Explaining the voluntary-involuntary contrasts.
Earlier in this chapter we suggested that any complete
theory of volitional control must explain the difference between
the voluntary-involuntary contrasts: similar-seeming pairs of
actions that differ only in that one is experienced as voluntary
while the other is not (Table 7.1). Three categories of
contrasting facts were explored in detail: the case of slips, of
automaticity, and of psychopathology. Here we attempt to show how
the theory we have developed so far can handle these facts.
An involuntary action that to escape voluntary control. It
is often known to be wrong at the very moment it is carried out.
We may hit a tennis ball with the sinking feeling that it is
going awry, and yet our own psychological momentum may be
unstoppable. Or we may make a slip of the tongue that never would
be made if we only had a little more time to think (Dell, 1986;
Baars, in press). When we make an error out of ignorance or
incapacity we do not speak of involuntary errors, errors that we
know are errors, and that îïîïwould have been avoided îexcept forï ---
what? One plausible explanation is that involuntary errors
involve a failure of anticipatory editing, as described above.
Editing occurs when systems that have spotted a flaw in a
conscious goal begin to compete for global access to keep the
goal from executing; but this editing function fails to work in
the case of slips, unwanted automaticity, and the persistent
errors of psychopathology. How could this happen?
Consider how editing might fail in our three primary cases:
slips, automaticity, and psychopathology.
7.51 Slips: a losing horse-race between errors and editing.
If conscious goal images tend to be carried out by default
when there are no competing elements, and if editing systems need
time to compete effectively against faulty goal images, there
must be a horse-race between "execution time" and "editing time."
("Execution time" can be defined as the time from the onset of
the conscious goal image to the start of the action. Similarly,
"editing time" is the time from the start of the goal image to
the beginning of effective competition that stops execution of
the act. See Figure 7.51.) In the case of slips, the editing
systems lose the horse-race, because execution time is faster
than editing time. The faulty action executes before editorial
systems have a chance to compete against its goal image.
There is one obvious case where this may happen: we know
that practiced images fade from consciousness or become very
fleeting, and that highly practiced, predictable actions become
more efficient and less conscious. Pani's (1982) studies on the
automatization of images show this pattern. As we have discussed
before (1.aa), Pani showed that conscious access to visual images
used in solving a problem drops consistently with practice. In
terms of our model, we can suppose that images become globally
available for shorter and shorter periods of time, until finally
they are globally available so briefly that they îcan no longer be
reported, even though they continue to trigger highly-prepared
effector systemsï. Highly prepared processors presumably can react
very quickly, while the act of reporting goal images may take
more time. Alternatively, it is possible that goal images are
simply lost from the global workspace; that they are not even
fleetingly available. In the remainder of this discussion I will
assume that the first case is true --- that with practice, goal
images are still broadcast globally, but more and more
fleetingly. Naturally this hypothesis must be tested (see 7.xx).
If goal images become more and more briefly available with
practice, the previously discussed studies by Langer and Imber
(1979) begin to make sense. These authors found that more
practiced subjects in a coding task were îmoreï willing to accept
an incorrect assessment of their own performance than less
practiced subjects. These authors argue that overlearning a task
can reduce the knowledge the subject has about îhowï the task is
performed, and under these circumstances subjects should be more
vulnerable to negative assessments of their own performance,
because they cannot evaluate their performance anymore by
themselves. This is exactly what we would expect, given the
assumption above that the goal image becomes less and less
available with practice. Automatic, highly prepared effector
systems can continue to carry out the task (because they became
more well-prepared and efficient with practice, and therefore
needed less of a goal image to be triggered). But asking someone
to do something novel such as evaluating their own performance,
should become more difficult, because the global goal image upon
with this evaluation can operate is available only fleetingly.
Thus the goal image controlling the "counter©voluntary" act
may be available long enough to trigger a îpreparedï action, but
not long enough to be vulnerable to interference from editing
systems.
In Figure 7.xx we show our usual model, with the goal image
G able to trigger off processors that tend to carry out goal G,
barring competing messages from other systems that may not
approve of G, which we will call ~G ("not-G") messages. If G is
globally available only very fleetingly, but long enough to
trigger well-prepared processors, then editing may fail because
the effectors may be faster than the editing systems. Systems
like ~G in Figure 7.xx find it difficult to interrupt, and modify
the goal image G. In this way, an action may "slip out" in an‹j ‹
uncontrolled way, because competing processors could not catch it
in time. Goal image G can come and go very rapidly, because there
are automatic systems able to execute it, and competing ~G
messages are too slow to stop its execution.
Notice a very significant point here: there is a trade-off
between îcompeting against Gï and îrepairing Gï. In order to correct
G, to modify it, to suggest alternatives, and the like, it is
important for many processors to have global access to it. G must
be available for a fairly long time if it is to be modified by
other systems. This is of course the whole point of making
something conscious, that many different unconscious experts can
cooperatively work on it (2.xx). But ~G systems compete against G
in order to stop its execution, and therefore make it îlessï
consciously available. If it is less available, there is less
time to modify G, and to improve it. This trade-off will be very
important in our discussion of psychopathology below.
Figure 7.xx tells why faulty goal images may be carried out
in spite of the fact that their faultiness is known; but it does
not tell us why the inner error occurred in the first place. In
the case of slips, I have argued in related work that competing
goals are often the cause of errors (Baars, 1980; 1985). For
example, there are often two different ways to express a single
thought. The two alternative goal images may compete for global
access, they may fuse or alternate. When there is a limited time
to resolve this goal competition, errors are likely to occur,
especially if other events load limited capacity at the same time
(Baars, in press, d; Chen & Baars, in press; Dell, 1986).
The horse-race between execution time and editing time is
key to the view of involuntary action we will maintain in this
discussion. It has strong implications not only for understanding
slips of the tongue, but also for unwanted automaticity and
psychopathology.
7.52 Counter-voluntary automaticity: the case of
"structural" slips.
Once the triggering conditions for any automatic process are
provided it becomes difficult to stop voluntarily. Habitual
cigarette smoking has an involuntary quality, as do compulsive
eating, nervous movements, and the like. Once we simply look at a
word, it is essentially impossible to stop reading it. The large
experimental literature on these phenomena makes the same point
(Shiffrin & Schneider, 1977; LaBerge, 1984). Habit is indeed the
"great flywheel of society," in James' well©known phrase, and
there are times when the flywheel runs out of control, resulting
in fatal accidents (Reason, 1984). Whenever we try to resist an
automatic habit, it will start to behave "erroneously" with‹j ‹
respect to our purpose. Such "errors" have much in common with
the slips discussed above. Just as in the case of slips,
automatic execution time is plausibly faster than voluntary
editing time. Thus we can apply the same "losing horse©race"
model in the case of unwanted automatisms. They seem to reflect
the same mechanism.
Of course, automatisms are not immune to change. Changing
them often requires repeated efforts. It may often be helpful to
block or slow down the action to make it more conscious and
easier to edit and modify. To speed up the editing process, we
may we need repeated trials to improve automatic editing and
re-establish voluntary control.
7.53 Psychopathology: the case of repeated editing failure,
perhaps due to excessive control effort.
If we are going in the right direction in this discussion,
what can we say about repeated errors that are îknownï to be errors
--- the case of psychopathology? The voluntary system we have
explored so far aims above all to achieve goals and minimize
errors. But in psychopathology we find a great range of behaviors
that violate voluntary goals, and that repeat errors with
remarkable persistence. Psychopathology seems to involve an
enduring failure of the entire voluntary control system. How
could such persistent failures arise?
Above, the lack of conscious availability was thought to be
responsible for a loss of editing ability. We block and fix
errors by making them available longer. If there is a repeated
editing failure in pathological symptoms, what could stand in the
way of this normal editing process?
One answer may be that îthe very attempt to block wrong goal
images may stand in the way of adaptation to the errorï. We have
referred above to the trade-off between modifying a goal îGï and
blocking its execution. That is, if we block a goal image, we
stop the goal from executing, but we also lose the opportunity to
modify and improve it. In order to fix a faulty goal image, we
must be allow it to be conscious for some time. But in the case
of pathological errors, editing systems may attempt to wipe the
goal image from consciousness as quickly as possible. In
psychopathology we may be trying to block the faulty goal image
so quickly and completely that we have no time to fix the
problem.
Take the example of a fearful image of an airplane crash.
Every time we think about taking an airplane trip, we may have a
vivid fearful image of the plane going down in flames. If we
allow ourselves to contemplate the image for a while, we may
notice that we can also mentally reverse the plane crash ©©© its
flaming wreckage may turn imaginatively into a whole new‹j ‹
airplane, and leap back up into the sky to continue its journey.
Just by allowing the image to remain conscious, many unconscious
processors will have access to the image. These unconscious
processors may be able to modify the conscious image in various
ways, thus creating a greater sense of control (Singer,
1984). The problem may come when we do not allow ourselves to
contemplate the fearful image at leisure. Rather, we edit it
quickly so as not to deal with its awfulness (Beck, 1976; Ellis,
1962). In that case, we do not provide the time needed to change
the image, to create alternatives, and the like. Then the fearful
mental image may become a rapid, frightening, and uncontrollable
phobic thought. It is this trade-off between "editing by
competition" and "fixing by conscious exposure" that may cause
phobic images to take on a life of their own.
If that is true, then allowing the phobic image to become
fully conscious, changing it to a safer image, and in general,
gaining more voluntary control over it --- all these methods
should work in the control of phobia. And indeed, these
techniques are the essence of phobic control: systematic
desensitization, imagery techniques, flooding, and practicing the
symptom may all work by allowing the phobic image to remain
conscious long enough to notice that the reality is not as awful
as the anticipation.
From this point of view the "paradoxical" techniques that
are sometimes so effective take on great importance. Voluntarily
getting children to stutter apparently solves the problem in some
cases; asking phobics to practice fearful imagery may help that
problem, and so on. These results make perfect sense from our
perspective: voluntary stuttering presumably causes a goal image
to remain conscious for a longer time, without destructive
competition to reduce its duration. And if it is available
longer, other systems can act upon the goal image to modify it,
so that it comes under the control of systems which failed to
control it before. Paradoxical practice of the to-be-avoided
action increases our ability to avoid the action.
It would be foolhardy to claim that this is the only
mechanism of psychopathology. But it may be one central factor
that sustains and aggravates a variety of repetitive
dysfunctional behaviors. It has great simplicity, there is some
good evidence for it, it is quite testable, and it flows
naturally from our entire discussion in this chapter.
In summary, we have explained the contrastive facts shown in
Table 7.1 by means of a modern ideomotor theory. It is plausible
that voluntary control is guided by momentary goal images, even
though those images are difficult to assess directly. The five
major points of the ideomotor theory seem to have some empirical
support, though more is needed. There is a satisfying fit
between the ideomotor theory and the theoretical approach to
consciousness we have been pursuing throughout this book. As we
see next, the ideomotor theory seems to generate fruitful‹j ‹
hypotheses about a number of problems, including the nature of
decision©making, perceived effort and control, the nature of non™qualitative
conscious contents, and even the understanding of
absorbed states of mind and hypnosis.
7.6 Wider implications.
7.61 What does it mean to make a decision?
Most of our actions are bound by past decisions that are not
currently conscious. As children we learned to pronounce the
difficult phoneme cluster /ask/ as "ask" rather than "aks," with
a lot of conscious concern for the different sounds. Once
learned, the difficulty fades into the background, and we need
not make the same decision again. All actions contain the residue
of commitments made at previous conscious choice-points,
decisions that are no longer conscious. If the goal hierarchy has
an established commitment to a certain decision, there is no need
to become conscious of the excluded alternatives. On those
potential choices we now have established policies.
But perhaps some aspect of almost any action is consciously
decided --- its timing, its propriety in a particular situation,
etc. Much of the time people can make voluntary decisions about
consciously entertained choices. We can decide to read a chapter
in this book, to adopt certain life-choices in adolescence, and
occasionally we can even make clear and effective decisions to
stop or start long-term habits. These are all choices with
conscious alternatives. If consciousness is the domain of
competition between such alternative goals, our model should be
able to show how we make decisions that stick, as well as those
that do not last.
The simplest approach is to say that one can broadcast
alternative goals, like "Should I îGï ... ?" followed by "Or
shouldn't I îGï ...?" and allow a coalition of systems to build up
in support of either alternative, as if they were voting one way
or another (Figure 7.61). The stronger coalition presumably
supports a goal image that excludes effective competition, and
which therefore gains ideomotor control over the action (7.0).
Thus voluntary actions may be preceded by a long set of problem-
solving triads, as described in Chapter 6.
----------------------------------------
Insert Figure 7.61 about here.
----------------------------------------
But where does the conscious goal image come from in the
first place? If the goal hierarchy is not strongly violated, it‹j ‹
presumably does not generate conscious goals (4.0). In that case
the hierarchy may îconstrainï the goals that are to become
conscious, without producing its own conscious events. Sometimes
of course the goal hierarchy is deeply challenged, and must
generate conscious goal images to maintain its integrity, or to
prepare for change (x.xx). Further, some conscious choices are
presented by the outside world, as when someone offers us a
tempting dessert, an attractive item on sale, or a career
opportunity. Other conscious choices are surely created by
internal changes, like the beginning of hunger or the onset of
puberty. Some may be created by unresolved conflicts between
deep goal structures, like the need to control others vs. a
desire to be liked by them. And some conscious choices may be
generated by a continuous process of entertaining long-term
dilemmas that have no simple solution.
All these points raise the issue of îindecisivenessï. As James
knew so well, the question of getting out of bed on a cold
morning appears as a struggle between alternatives. Perhaps most
of our ordinary decisions have this quality. But some extended
struggles may be won by patience rather than force. As James
noted in the epigraph, one can simply wait until the cold of the
morning fades from consciousness; if a fortuitous thought about
getting up then emerges, it may be able to dominate consciousness
without competition. Thus the persistently returning thought
ultimately wins out. The idea that important goal systems can
"win out" by sheer persistence, by returning to consciousness
again and again, is consistent with evidence from thought
monitoring (Klinger, 1971; Pope & Singer, 1978), showing that
unresolved issues tend to recur spontaneously.
Indecisiveness may be the case where neither of the two
contending goals ever completely fades away. Young children often
seem indecisive compared to adults. They may be quite impulse-
driven, sometimes hesitating back and forth, back and forth,
between two attractive goals. Young children may not yet have a
consistent dominant goal hierarchy. Over time, many repeated
cases of effective coalition-building between the most successful
goals may result in the relatively stable adult goal-hierarchy,
so that consistent goal contexts become established and
automatized like any other skill.
We have noted that conscious goals that are consistent with
the goal hierarchy will last longer than those that violate it
(7.83). It also follows from our current model that some
conscious goal images may fit the deeper levels of the goal
hierarchy better than the more superficial levels. In that case
the more deeply- driven goal images may last longer, or they may
return again and again until they lead to action. One way to make
new goals effective is to tie them in with existing deep goals.
Thus one may have an inner argument of the form: does my
commitment to survival make it necessary to go out and jog four
miles a day? Does my commitment to social success make it
imperative to stay at this boring party? In these cases a‹j ‹
conscious connection is created between an immediate goal and an
existing deep commitment; in just this way politicians will make
a case for new expenditures by reference to existing deep and
agreed-upon goals like "national security," "winning the war on
poverty," and "bringing back prosperity." By consciously
mobilizing the deep goal hierarchy, one's superficial reluctance
may be overcome. These rhetorical connections between local goals
and deep goals may be specious, but as long as they allow the
conscious goal to be available long enough to be executed, they
will be effective.
Below we argue that most normal action is relatively
conflict-free (section 7.83); that is, it takes place in the
domain of overlap between many deep goals. But conflicting goals
are not unusual. Any new major goal must of course be reconciled
with existing priorities.
Thus much of the time people may carry on an inner argument
about their goals. Not all of this inner argument may be fully
conscious; some of it may consist of fleeting images that
function merely as reminders. Notice an interesting thing about
this inner argument: If the ideomotor theory is true, it is very
important to îhave the last wordï in a train of arguments; the last
word, after all, is the one that will be carried out, because it
is not followed by competition.
Inner arguments about goals have many implications. For
example, one can define a "belief" as an abstract concept that is
not disputed in the stream of thought, though it could be. One
can dispute a political or religious belief, or a scientific
position, but it is quite remarkable how rarely people challenge
their own beliefs. A belief îsystemï may be defined as a consistent
set of such undisputed concepts, one that serves to stabilize
one's thoughts, feelings, and actions --- presumably just by
giving the belief system the last word in the inner argument,
since the last word has the real power by the ideomotor theory
--- it is the one that controls action without contradiction.
Likewise, a îclosedï belief system is one that has a ready answer
to all apparent counter-arguments, so that any possibility of
change is minimized (Adorno, et al, 19xx; Rokeach, 1960). Perhaps
all ideological, political, philosophical, and even scientific
belief systems are closed to some extent. Simple observation
should convince us that most people have self-serving closed
belief systems about many disputable topics, especially those
that are difficult to decide on direct evidence.
‹` ‹
7.62 Resistance to intended actions, perceived effort, and
perceived voluntary control.
We noted above that voluntary control is different from
spontaneous problem solving (6.0) in that we usually îknowï that
our voluntary actions are goal-directed. We have "metacognitive"
access to many voluntary goals, and often to mental events that
block voluntary goals (e.g., Flavell & Wellman, 1977). Two
factors may give us this kind of metacognitive access. The first
is obvious: îex hypothesiï, the ideomotor theory states that
voluntary control involves conscious images, which, if they are
available long enough, are also available to metacognitive
processors. Metacognitive processors are presumably involved in
representing, recalling, and describing the fact that we do have
a certain conscious goal. Thus the reader knows that s/he is
reading this book voluntarily, in part because he or she may be
able to recall the conscious goal of doing that.
But there must be many times when we experience an action as
voluntary even when we do not remember its controlling goal
image. After all, goal images are fleeting, their memory may be
masked by later events, and so on. By our discussion above, the
more automatic the action, the less we can report our intention.
Conversely, the more the action îencounters resistanceï, the less
automaticity will operate, and the more a decision to act can
typically be reported.
This suggests that resistance to performing an action, the
perception of effort, and perceived voluntary control are all of
a piece. Take the issue of perceived effort, which appears very
much as a conflict between expected control and actual control.
The author's experience in typing the manuscript of this book may
illustrate the point. As a practiced (though errorful) typist, I
am normally unconscious of the details of typing. The computer
program that displays words on my screen works so quickly that
normally I do not notice it at all. But sometimes when I am
typing, the computer is simultaneously printing out some other
material, and then the screen seems to slow down enormously. The
lag time between a finger stroke and a character appearing on the
screen is now very long, compared to my expectations. Now the
relationship between key-strokes and characters on the screen
becomes agonizingly conscious. The subjective experience is one
of great effort, as if I must forcibly push each character onto
the screen. I am acutely aware of the voluntary character of
every keystroke.
This example may provide a way to test the hypothesis that
perceived voluntary control results from perceived effort. It
suggests that a goal context contains information about the
length of time an action should take. When this time is delayed,
we tend once more to become conscious of both goal images and
feedback, so that many processors can now operate on the
conscious components of the action. As the conscious goal image‹j ‹
becomes more available, metacognitive processors can also operate
on it, to facilitate recall and self-description. In this way,
our knowledge that we have a certain goal may depend on
violations of automaticity in accomplishing the goal.
Notice that the increase in conscious access to such a
delayed goal gives us four distinct advantages: first, we have
more time to edit and change the conscious goal; second, in this
process we improve our voluntary control over the action; third,
we can comment on the goal in question metacognitively, which
then allows us to recall it, to talk about it, and perhaps to
find alternative ways to accomplish the same ultimate end.
Finally, as we will see in Chapter 8, access to a conscious goal
can also guide îlaterï conscious contents, as when we make
conscious decisions about what to pay attention to next.
7.63 Ideomotor control of conceptual thought: A solution to the
puzzle of non-qualitative consciousness?
One of our persistent thorny problems has been the
relationship between clear, qualitative conscious contents like
percepts, feelings and images îversusï non-qualitative conscious
events like concepts, beliefs, expectations, and intentions,
which surely take up limited capacity, but are not experienced
with qualities like warmth, color, taste, and smell (x.xx).
We have previously remarked on the fact that human beings
have a great tendency to concretize abstract ideas: to think in
terms of metaphors that can be visualized, or to reduce an
abstract class of events to a concrete prototype. It may be no
accident that mathematics and physics really have two separate
symbol systems: an algebraic code and a geometric one. The two
are mathematically equivalent, but not psychologically, because
people can use their visual imagination with geometric graphs but
not with algebraic formulas.
But we do not have to resort to science for examples. All of
us clearly represent the meaning of a sentence in an abstract
form. To illustrate this, let the reader recall word-for-word the
sentence before this one. (No looking!). The great majority of
readers will not be able to do this, but they will be able to
recall a îparaphraseï of the sentence --- that is to say, a
semantic equivalent, with different words, different syntax, and
even different sensory qualities than the original sentence; but
the paraphrase will preserve the abstract meaning of the
original. The evidence is very good that educated adults rapidly
convert words and sentences into a semantic code that is quite
abstract and impossible to experience qualitatively (Bransford &
Franks, 1976). The question we must face here is, of course: How‹j ‹
do we then manipulate the abstract semantic code through
consciousness?
One possibility is that we have ideomotor control over
abstractions. Take the complex conceptual structures developed in
the course of this book. We have now defined îtermsï like
"context," "global workspace," and the like, which are perceptual
in nature even though they refer to abstract non-qualitative
things. The more we can manipulate these abstractions with words
we can see and hear, the easier we will find it to understand the
theory. Likewise, we have made a great effort in this book to
present useful metaphors for our theory, such as the image of a
conference of experts, each competing for access to a central
blackboard. But every concrete metaphor is really inadequate. The
conference image fails to show that expert processors in GW
theory are decomposable, while human experts are not (2.xx).
(Human experts have a tendency to stop running when they are
decomposed.) This point is quite general: metaphors are
inadequate representations of a more abstract and complex reality
(Lakoff & Johnson, 1980). In science, they must be used with
care.
In general, an imageable metaphor seems to serve the
function of evoking and recruiting conceptual processes that are
more abstract and often more accurate than the image itself.
These abstract entities may be impossible to experience
qualitatively. Hence the need for visual figures, audible words,
and concrete metaphors. These can be made qualitatively conscious
when needed, to stand for abstract non-qualitative entities.
All this suggests that we do indeed have ideomotor control
over abstract concepts, so that we can always concretize an
abstraction, and conversely, we can always abstract from concrete
symbols. It is not enough merely to translate the perceptual
world into abstractions as we do in comprehending a sentence; in
turn, we must be able to retrieve the abstractions in perceptual
form in order to work with them, to resolve conflicts between
them, to make predictions from them, and to use them to act on
the world. In all these transformations, it is useful to recode
the abstractions into some qualitative, imageable form. The
ideomotor theory seems to add some real clarity to the
problematic relationship between qualitative experience and
abstract representation.
7.64 Fleeting goal images make accurate source attribution
difficult.
If it takes time for a goal image to result in action, then
what about goal images that not only trigger an action, but also
require us to talk about them? That is, images that recruit îtwoï
actions? If we want people to report their own goal images, they‹j ‹
must make the goal image available long enough not only to
trigger the original action, but also to help recruit linguistic
systems able to describe the image. This is of course the same
problem we encountered before, when we first raised the
possibility of fleeting conscious events that pass too fast to
describe (1.xx). The best example, again, is in tip-of-the-tongue
states when people experience the missing word fleetingly, and
encounter the frustration of trying to hold on to the image long
enough to say it.
This is a fundamental problem in metacognition. It may help
to explain a number of problems in self-observation. There is of
course the social©psychological literature on errors in
attribution of personal causation, and the frequent failure of
people to know their own reasons for doing things (Nisbett &
Wilson, 1977; Weiner, 19xx). Accurate source attribution is very
important for metacognitive knowledge and self-control. Again,
this is not a topic we can explore in much detail, but we can
suggest its relevance to GW theory.
7.65 The relationship between a goal image and the action it
evokes may be highly variable.
The act of walking to the kitchen may be triggered by many
different goal images. One can imagine a seductive peanut-butter-
and-jelly sandwich, or left-overs from last night's dinner; one
can remember that the stove needs cleaning, or imagine the odor
of cooking gas. We need not imagine any of these in great detail.
A fragment of a related image will do quite nicely to trigger a
habitual action. This is very much like the issue of synonymy and
paraphrase in language: there are dozens of ways of saying the
same thing. In action control, a conscious cue is presumably
interpreted by many different context-sensitive systems. We do
not need a detailed conscious plan or command, since the action
is carried out by specialists that know more about local
conditions than we do consciously. Various unconscious
specialists keep continuous track of our posture, balance and
gravity, about salivation and digestive enzymes to prepare for
eating, about remembering the route to the kitchen. Greene (1972)
has pointed to the ambiguity of commands in distributed control
systems as a general and very useful property.
This point has important implications for research. We must
not fall into the trap of looking for îtheï goal image for walking,
or talking, or for any other action that looks the same in
different circumstances. This is what misled introspectionists
like Titchener and Kuô"lpe, who were astonished to find the great
range of variation in mental images between different observers‹j ‹
(1.xx). The modern ideomotor theory indicates that many different
goal images can serve to recruit and initiate any given action.
Conscious images may seem quite irrelevant, and still result in
appropriate action. Imagining a sandwich while walking in the
desert must not trigger an automatic walk to the kitchen, but it
can stimulate new efforts to find food and water. Thus goal
images may vary tremendously between different situations and
observers, and yet be quite effective in controlling normal
voluntary action.
7.7 Absorption and hypnosis as ideomotor events.
7.71 Absorption as a drop in competition for GW access.
The ideomotor theory has many interesting implications. For
example, it suggests a reasonable account of hypnosis as a state
in which ideomotor control operates without effective competition
(7.67). Before we discuss this, we can define îan absorbed stateï
--- watching a movie, reading a novel, and the like --- as a
state in which only one coherent stream of events dominates
consciousness (viz., Tellegen & Atkinson, 1974; Spiegel, 1984).
That is, there is a low level of effective competition between
different topics (dominant contexts), and there is no voluntary
effort to change topics (see Chapter 8).
In principle, it would seem that there are two ways to reach
an absorbed state. One is for the number of competing contexts to
decrease. This may happen simply when we relax, let go of our
current concerns, solve a major preoccupying problem, or enter a
state of trust that things will work out without voluntary effort
(Klinger, 1971). A second way to enter an aborbed state is to
allow one context to become extremely dominant and thereby to
exclude alternatives. Shadowing tasks compel one to repeat
immediately each word in a coherent stream of words (Cherry,
1953; Broadbent, 1958). This task is so demanding that competing
thoughts are simply excluded from conscious experience.
Nevertheless, competing thoughts have probably not disappeared;
they are simply excluded from consciousness. Thus we can enter an
absorbed state either if consciousness is dominated by a very
strong context, or if there is a drop in competition from
alternative contexts. In fact, of course, most actual absorbed
states have both of these features. In watching a fascinating
movie our experience is being structured by the story line, which
continually generates new expectations about future events that
need to be tested. At the same time we may relax, postpone some
pressing concerns, and thus lower the urgency of competing
topics.
One implication is that îwe are always in an absorbed state
relative to our own dominant contextï. If we look at the goal
hierarchy (x.xx), we can see that its lower levels can change‹j ‹
much more easily than higher goals, which are quite stable over
time. Most people do not cease wanting to survive, to be socially
accepted and respected, and to pursue other life-long goals.
Adults change their major beliefs and goals quite slowly if at
all. Even perceptual and imaginal contexts change only rarely.
This suggests that we are never absolutely "absent-minded,"
"mindless," or even "preoccupied" (Reason & Mycielska, 1984). We
are always "present-minded" to our îownï dominant preoccupations.
Now, if we are driving a car and thinking thoughts of love at the
same time, we may run over a pedestrian. îRelative toï the act of
driving the car we were preoccupied and absent-minded. But
relative to thinking thoughts of love, we were quite present.
Taking a bird's eye view of the situation, it would seem
impossible to be utterly absent-minded. "Absorption" is only a
relative term.
When we are absorbed in one mental topic to the exclusion of
others, the other topics must go on automatic. Thus if we were to
ask someone to shadow speech while performing a fairly routine
task --- driving a car along a familiar route --- we would see
the automatic components of driving emerge with minimal conscious
and voluntary overlay. We should then expect to find large
numbers of automatic "habit intrusions" into the act of driving
(Reason, 1983). Driving a car distractedly may be rather
suicidal, but similar experiments can be done under less
dangerous circumstances.
7.72 Hypnosis as ideomotor control without competition.
Absorption has long been thought to be a key element in
hypnosis. When we combine the idea of absorption with ideomotor
control, we have a possible theory of hypnosis (James, 1890, ch.
on hypnosis). The major features of hypnosis seem to flow from
the fact that in this state we have only one conscious goal image
at a time, which tends to be carried out because the chances of
competition from other elements are reduced. Although we cannot
go into hypnosis in great detail, this possibility is worth
exploring briefly.
What are the major features of hypnosis? There seems to be
good agreement on the following:
1. îAbsorptionï, sometimes called "monoideism," or
"imaginative involvement" (Tellegen & Atkinson, 1974; Spiegel &
Spiegel, 1978; J. Hilgard, 1979; E. Hilgard, 1977; Ellenberger,
1970). Hypnosis seems to create a new, imaginative context that
dominates experience for some time to the exclusion of other
events (Singer, 1984).
‹j ‹å 2. îDissociationï. Good hypnotic subjects show several kinds
of spontaneous dissociation. First, there are two kinds of
temporal dissociation. A good subject is often spontaneously
amnesic for the experience, which is a kind of post-hypnotic
temporal dissociation. There is also pre-hypnotic dissociation,
since separation from previously dominant trains of thought is
common (J. Singer, personal comm.). In addition to temporal
dissociation, two kinds of îconcurrentï dissociation occur. These
may be called "dissociation from effectors" and "dissociation
From the normal self." Subjects often report feelings
alienation from their own limbs that are manipulated by
suggestion, as if their arms and legs had "a will of their own"
(Spiegel & Spiegel, 1978). Further, there is commonly some
surprise at îoneselfï for allowing the hypnotic actions and
experiences to happen, so that there is a kind of dissociation
between one's "normal self" and one's "hypnotic self" (viz.,
9.x). In sum, there is an experienced îtemporal separationï from
earlier and later states, and also îconcurrent separationï during
hypnosis from one's own normal experience of self and one's own
hypnotically controlled actions.
3. îSuggestibilityï is a defining feature of hypnosis; but
this can be viewed as a consequence of ideomotor control, plus a
kind of dissociation from normal inhibitions. But dissociation
From inhibiting thoughts is of course a property of absorptio
as a state of low competition for consciousness. If the ideomotor
theory is true, and if our normal editing mechanisms are not
competing against novel conscious contents, it follows that one
will show a great flexibility in obeying the conscious ideas.
Thus suggestibility seems to flow from "absorption plus ideomotor
control."
4. Strong and stable îindividual differencesï. About a quarter
of the population is highly hypnotizable. These people easily
slip into this remarkable state with a very simple, standard
induction (Hilgard, 1977; Spiegel and Spiegel, 1978).
5. Hypnotic îinductions are arbitraryï: anything believed by
the subject to induce hypnosis will induce hypnosis (James,
1890). However, relaxation and a reasonable feeling of trust for
the hypnotist are common features.
These are the positive facts about hypnosis that command a
good consensus. At the same time hypnosis has some îpuzzling
"negative" featuresï: properties that we might expect, but which
researchers have not found in spite of repeated efforts.
1. No reliable neural correlates of hypnosis have been found
so far. Physiologically, hypnosis looks like a normal waking
state.
2. There is a puzzling absence of strong and reliable
personality correlates, in spite of many attempts to find them
(e.g., J. Hilgard, 1979). ‹j ‹å
3. It is difficult to demonstrate a conclusive difference
between hypnosis and "pretense" (T.X. Barber, 19xx). But this may
be in good part because very good "pretenders" are able to
experience their pretended states very deeply and realistically
--- in other words, good pretenders may enter an absorbed state,
in which only one train of conscious contents dominates their
actions. There may thus be no real difference between very good
play-acting and hypnosis, but this fact may reveal as much about
acting as it does about hypnosis. Many professional actors
experience deep absorption and identification with the characters
they play. The difference between half-hearted acting and
Stanislawskian "method acting" is the difference between being
superficially involved, and being deeply absorbed in a certain
character (Stanislawsky, 19xx). Stanislawskian method actors may
believe for a while that they are the character they are playing.
Absorption may be the key both to good hypnosis, and to good
pretense as well.
îHypnosis as absorbed ideomotor control.ï
Several investigators maintain that absorption may be the
basic element of hypnosis, the single central fact from which all
else flows (e.g. Spiegel and his refs). This is a very attractive
argument from our point of view. We can simply take James'
ideomotor theory, in our modern version, and explain all of the
features listed above. That is:
1. îAbsorptionï or monoideism is simply a low level of
competition for access to consciousness between alternative
contexts. Under these conditions of "low editing" the dominant
stream of consciousness, which may be quite different from our
normal states, should be in control. *** ALSO metacognition, see
next Chapter.
2. îDissociation.ï We can easily explain both temporal and
concurrent dissociation. Spontaneous amnesia after hypnosis is a
difficulty in voluntarily reconstructing the hypnotic state in
such a way as to easily retrieve information from it. This is not
surprising, given the differences in content between the
hypnotic, absorbed context and our more usual contexts, those
which we îcallï normal and not hypnotic. So spontaneous amnesia
would seem to follow quite easily. The same argument applies to
the temporal dissociation from the dominant context before
hypnosis. We should feel dissociated from it, given the
differences in content. If hypnosis is mainly an absorbed state,
there should be relatively few shared features between it and our
normal state, thus making recall difficult.
The two kinds of concurrent dissociation also make sense.
Dissociation from our own hypnotically controlled limbs may be‹j ‹
just the act of noticing the truth of ideomotor control. In our
normal waking state we tend to forget that are îalwaysï unconscious
in detail of actions that carry out our conscious goals. That is
what the ideomotor theory is all about, after all. Similarly, we
are normally unconscious of the controlling contexts of our own
actions. In hypnosis we may be surprised to realize that. But in
our whole approach in this chapter, the unconsciousness of goal
contexts and automatic actions has become a fundamental
assumption about normal action. From this point of view, it is
not dissociation that is unusual. What is novel in hypnosis is
the fact that we îrealizeï the existence of ideomotor dissociation
between conscious events, their consequent actions, and their
antecedent goal contexts. Perhaps we recognize this in hypnosis
because hypnotically controlled actions are often unexpected;
they violate our usual contextual assumptions about ourselves.
The close connection between absorption and dissociation
comes out in Spiegel's (1984) clinical observation that
"... it has been commonly observed that many highly
hypnotizable performers, such as actresses and musicians,
dissociate their ordinary awareness of themselves when they are
performing, and feel strangely disconnected from the results of
their peformance after it is over. One highly hypnotizable
pianist reported that her only memory of her graduation recital
was of floating above the piano admiring the grain of the wood.
She had to ask a stagehand whether she had in fact completed her
program, which she had performed extremely well."
Other features of hypnosis are also consistent with this
point of view.
3. îSuggestibilityï and flexibility seem to be merely the
result of ideomotor control in an absorbed state, with minimal
competition and self-examination.
4. The reasons for the strong and stable îindividual
differencesï in hypnotizability are not clear. Given that the
notion of hypnosis as "absorbed ideomotor control" seems to work
quite well, the question seems worth exploring from this
theoretical perspective.
5. The îarbitrarinessï of hypnotic induction techniques is
quite understandable, since we know that any conscious experience
may trigger a context (4.xx). Hypnosis involves a context, one of
minimal competition for GW access. Relaxation and trust for the
hypnotist may be simply different ways of describing this
absorbed state.
We can also make sense of some of the negative results, the
absence of expected features of hypnosis. Hypnosis appears to be
such a spectacularly different state of mind that many‹j ‹
researchers expected to find major physiological and personality
differences. But if we assume that hypnosis is not an unusual
state at all, but is rather a state of low competition for access
to consciousness, we should find no physiological differences
between hypnosis and relaxation. The absence of personality
correlates is not surprising either, because we are all absorbed
in our own top-most goal context, as suggested in section 7.71.
In that sense all personality types involve absorption.
Finally, we should find it hard to distinguish between hypnosis
and very good pretense, because successful pretense is like
excellent performance in any other demanding, complex task. It
requires absorption.
In sum, hypnosis may simply be ideomotor control in a state
of absorption. But absorbed states are quite normal, and in a
general sense, we are all absorbed in our own top-level contexts.
The major difference seems to be that highly hypnotizable
subjects are quite flexible in the topics of their absorption,
while most people are not. Perhaps we should turn the usual
question around. Instead of asking, what is îdifferentï about
hypnosis? we might ask, why is flexible absorption so difficult
for three-quarters of the population? What is it that is added to
a "ground state" of absorption, which we all share, that resists
flexible ideomotor control? We will explore this question in the
next few chapters.
7.8 Conflicts between goals.
We have already discussed the possibility of competing goals
and contexts (4.xx); here we will explore the implications for
conflicting emotions. Goals can encounter conflict, either from
other goals or from reality. All emotions involve goals, combined
with real events. Happiness may result from achieving a
wished-for goal; sadness involves loss of a desired object,
depression is due to helplessness and hopelessness about
significant life goals, anger and frustration occur when
obstacles stand in the way of achieving a desired goal, fear is
due to the expectation that something will happen that is
fervently desired înotï to happen, love involves the goal of being
with someone, and so on. All these goals can be represented in GW
theory. But all these emotions involve clear, dominant goals that
can be consciously achieved, delayed, thwarted, and the like.
The really difficult cases for voluntary control arise when
this is not true; when there is competition for access to
consciousness between different goals, so that no single goal can
dominate. We have already discussed indecision due to conflicts
between goals, and the possibility of an "inner argument," in
which the final word wins ideomotor control. William James'
discussion of "weakness of the will" and "explosive will" is also