î
Part V. Attention, self, and conscious self-monitoring. Ñ
ïIn everday language, the word îattentionï implies control of
access to consciousness, and we adopt this usage here. Attention
itself can be either voluntary or automatic. This can be readily
modeled in the theory. Further, a contrastive analysis of
spontaneously self©attributed vs. self©alien experiences suggests
that îselfï can be interpreted as the more enduring, higher levels
of the dominant context hierarchy, which create continuity over
the changing flow of events. Since context is by definition
unconscious in GW theory, self in this sense is thought to be
inherently unconscious as well. This proposal is consistent with
a great deal of objective evidence. However, aspects of self may
become known through îconscious self-monitoring,ï a process that îïis
useful for self-evaluation and self©control. The results of
conscious self-monitoring are combined with self©evaluation
criteria, presumably of social origin, to produce a stable îself™conceptï,
which functions as a supervisory system within the
larger self organization.
‹> ‹
Chapter Eight
Model 6: Attention as control of access to consciousness.
8.0 Introduction: attention versus consciousness.
8.01 Attention involves metacognition.
8.02 Attention is controlled by goals.
8.1 Voluntary and automatic control of access to consciousness.
8.11 Voluntary attention: Conscious control of access to
consciousness.
8.12 Automatic attention: unconscious control of access to
consciousness.
8.2 Modeling voluntary and automatic access control.
8.21 Model 6A: Automatic attention is guided by goals.
8.22 The Options Context for conscious metacognition.
8.23 Model 6B: Voluntary attention allows conscious choice.
8.24 Automatizing voluntary attention.
8.3 Directing attention îtowardï something.
8.31 Using voluntary attention to create new access
priorities.
8.32 Mental effort due to competition between voluntary and
automatic attention.
8.33 The importance of accurate source attribution.
8.4 Directing attention îaway fromï something: suppression,
repression, and emotional conflict.
8.41 Thought avoidance can occur in many ways.
8.42 When we lose voluntary access to avoided thoughts,
there is the appearance of repression.
8.43 Signal anxiety as ideomotor control of thought avoidance.
8.5 Further implications.
8.51 Absorption and suspension of disbelief.
8.52 Deep hypnosis may decrease access to options contexts. ‹j ‹å
8.53 The role of attention in motivation and maintenance of
mental stability.
8.54 Further thoughts on the operational definition of
conscious experience.
8.6 Summary.
î
ï‹
8.0 Introduction.
Common sense makes a distinction between attention and
consciousness. In English we can ask someone to "please pay
attention" to something, but not to "please be conscious" of it.
But of course, when people pay attention to something they do
become conscious of it. Likewise, we can "draw," "get," or "call"
someone's attention îinvoluntarilyï by shouting at, waving to, or
prodding the person; as soon as we succeed, our subject becomes
conscious of us. Nevertheless, we still do not speak of "getting"
or "drawing" someone's consciousness.
It seems as if the psychology of common sense conceives of
attention as something more active than consciousness, while
consciousness itself is thought of as a "state." A similar
distinction is implied in pairs of verbs like îlookingï vs. îseeingï,
îlisteningï vs. îhearingï, îtouchingï vs. îfeelingï, etc. In each case
the primary sense of the first verb is more active, purposeful,
and attentional, while the second verb refers to the conscious
experience itself. Nor is this distinction limited to perception:
it also works for memory (as in îrecallingï vs. îrememberingï); even
in the case of imagination, the verb îimaginingï is more active and
purposeful than îdaydreamingï.
Of course consciousness is profoundly active, even when it
is not experienced as such. And we have previously suggested that
superficially purposeless thoughts may in fact serve specific
goals (6.0). But the commonsense distinction between attention
and consciousness is still important. It embodies the insight
that there are attentional control mechanisms for access to
consciousness --- both voluntary and automatic --- that determine
what will or will not become conscious. It implies that attention
involves metacognitive operations that guide the stream of
consciousness.
This belief is backed by good evidence. We can obviously
control what we will be conscious of in a number of voluntary
ways: we can decide whether or not to continue reading this book,
whether to turn on the television, whether to stop thinking some
unpleasant thought, and so on. Prototypically we can treat
attention as the control of eye©movements. We can decide
voluntarily to look at an object, or to look away from it; in
these cases, a conscious, voluntary, reportable decision
precedes the act of attending (see Table 8.0). But of course,
most eye movements are not consciously controlled; they are
automatically controlled by sophisticated systems that are not
normally reportable. Of course the control of eye movements is
only a convenient prototype of attention. There must be analogous
access©control systems for all the senses, for memory retrieval,
for imagery, knowledge, and indeed for all sources of conscious
contents. There is even attentional control in vision without
moving the eyes (Posner & Cohen, 1982). Thus attention as access‹j ‹
control is not one system but many; nonetheless, we will try to
describe common features of all access control systems.
In particular, the distinction between voluntary and
automatic control is crucial. Without flexible, voluntary access
control, human beings could not deal with unexpected emergencies
or opportunities. We would be unable to resist automatic
tendencies when they became outdated, or to change attentional
habits to take advantage of new opportunities. Without automatic
access control, rapid shifting to significant events would be
impossible. We need both voluntary and automatic attention.
For all these reasons we adopt here the commonsense view of
îattentionï as îthat which controls access to conscious experienceï.
This chapter aims to enrich GW theory with this capacity.
---------------------------------------------------------------
Table 8.0: Attentional contrasts.
---------------------------------------------------------------
îConsciousï îUnconsciousï
Voluntary decisions Automatic mechanisms for
to make things conscious. access to consciousness.
(e.g. Being asked to pay (e.g. One's own name
attention to something, breaking through from an
especially when this is unattended stream of speech.)
effortful.)
-----------------------------------------------------------------
8.01 Attention is metacognitive.
The control of access to consciousness is inherently
metacognitive. That is, it requires knowledge îaboutï our own
mental functioning, and about the material that is to be selected
or rejected. Voluntary attention would seem to require conscious
metacognition, or the ability to have conscious access to and
control over the different things that can become conscious.
Metacognition is a major topic in its own right, one we can
only touch on here. It is widely believed that knowledge of one's
own performance is required for successful learning (Flavell &
Wellman, 1977). Among students in school, good learners
continually monitor their own progress; poor learners seem to‹j ‹
avoid doing so, as if fearing that the results might be too awful
to contemplate (Bransford, 1979). But by avoiding conscious
knowledge of results, they lose the ability to guide their own
learning in the most effective way. We have previously
maintained that consciousness is involved especially in the
learning of new things, those that demand more adaptation
(Chapter 5.0). If that is so, then attention (as metacognitive
control of consciousness) seems to be necessary for voluntary,
îpurposefulï learning.
We have already remarked on the fact that we can only know
that someone is conscious of something by a metacognitive act: "I
just saw a banana." That is why our operational definition (1.xx)
is unavoidably metacognitive. But this is not the only important
link between metacognition and consciousness. Many of the most
important uses of consciousness are metacognitive. Normal access
to Short Term Memory involves metacognitive control of retrieval,
rehearsal, and report (see Figure 8.22). Long Term Memory
retrieval is equally metacognitive. One cannot know consciously
why or how one did something in the past without metacognitive
access. One cannot deliberately repeat an action by evoking its
controlling goal image (7.xx) without metacognition, nor can one
construct a reasonably accurate and acceptable self©concept
(9.xx) without extensive metacognitive operations. All these
functions require sophisticated, and partly conscious,
metacognitive access and control, which inevitably becomes a
major theme of this book from here on.
One of our main concerns in this chapter is the role of
conscious metacognition in voluntary attention. One cannot
choose consciously between two alternative conscious topics
without anticipating something about the alternatives. That is,
one must represent to oneself what is to be gained by watching
the football game on television rather than reading an
interesting book. Further, one must operate upon one's own system
in order to implement that voluntary choice. These are all
metacognitive operations. Later in this chapter we will propose
that human beings have access to something analogous to a
computer "directory" or a "menu," which we will call an Options
Context. The Options Context makes consciously available whatever
immediate choices there are to attend to in sensation, memory,
muscular control, imagery, and the like (Figure 8.22). Voluntary
control of attention then comes down to making a conscious
decision (7.xx) about current options. Since Options Contexts are
called through the global workspace, control of the global
workspace also determines one's ability to control voluntary
attention.
8.02 Attention is directed by goals.
We suggest in this chapter that attention is always‹j ‹
controlled by goals, whether voluntarily or not. This is obvious
in the case of voluntary attention: when we have the conscious,
reportable goal of paying attention to the weather rather than to
a good novel, we can usually do so. It is not so obvious for
automatic attentional mechanisms, but the same case can be made
there. If one's own name breaks through to consciousness from an
unattended stream of material, while a stranger's name will not,
this suggests that significant stimuli can exercise some sort of
priority; but significant events are of course those that engage
current goals (Moray, 1959). Since there was in this case no
conscious decision to pay attention, there must be unconscious
mechanisms able to decide what is important enough to become
conscious. One's own name must surely be associated with the
higher levels of the goal hierarchy. Indeed, we have previously
touched on the idea that significant input can be defined as
information that reduces uncertainty in a goal context; there is
thus an inherent relationship between significance and goals
(5.xx). Therefore it seems likely that automatic attention is
controlled by the goal hierarchy in such a way that more
significant events tend to out-compete less significant ones.
Automatic attentional mechanisms have been widely
investigated, in general by training subjects to detect or
retrieve some information; but of course training works by
teaching subjects that some previously trivial event is now a
significant goal in the experiment ©©© that is, in experimental
training we always transmute the social and personal goals of the
subject into experimental significance. A subject may come into
the experiment intending to cooperate, to appear intelligent, to
satisfy his or her curiosity, or to earn money; in the course of
the experiment, this translates into a good©faith effort to
detect tones in auditory noise, or to spot faces in a crowd (e.g.
Neisser, 1967). Successful performance on this previously
irrelevant task is now perceived to be a means to satisfy one's
more personal goals. Thus whatever motivates the subject to
participate now becomes the indirect goal of the experimental
task as well. In this sense, significance is something we create
by the very social conditions of an experiment. Any experiment
that trains automatic attention is therefore involves a
significant object of attention.
Thus attention has to do with the assignment of îaccess
priority ïto potentially conscious events. Practicing voluntary
attention to the point of automaticity is known to improve the
chances of an event becoming conscious (e.g. Neisser, 1967,
8.xx). But of course in the real world those events that we
decide voluntarily to pay attention to most often, and which
therefore become highly practiced, are precisely those that are
significant. We practice more voluntary attention to the color of
a traffic light than we do to the paint on the pole that holds up
the light. Thus in the real world, significance and practice
covary.
8.1 Voluntary and automatic control of access to
consciousness.
We will argue in this chapter that voluntary control of
attention may be quite flexible, in contrast to automatic
attention, which is relatively rigid because unconscious and
automatic processes are insensitive to context (2.xx). While
automatic attentional mechanisms seem to be controlled by the
goal hierarchy, there is reason to think that voluntary
attention can operate at least for to a degree independently.
Even the most compulsive dieter can voluntarily ignore, at least
for a while, the presence of delicious, tempting food. But there
is a trade©off between habitual, goal©guided control of
attention, and voluntary attention. We now explore these issues
in some detail.
8.11 Voluntary attention: conscious control of access to
consciousness.
According to the last chapter, volition comes down to
ideomotor control. That is, it involves a momentary conscious
goal image that serves to recruit unconscious processors needed
to carry out the goal. Thus voluntary control requires
consciousness, at least briefly. But attention, by the argument
above, is the control of îaccessï to consciousness. It follows that
voluntary attention is conscious control of access to
consciousness.
This may sound paradoxical, but it is not. We can be
conscious of the next thing we want to be conscious of, and
display a goal image to embody that intention. This goal image in
turn can trigger unconscious processes able to fulfill the
intention. An obvious everyday example is the intention to watch
a news program on television. Once the goal of watching the news
becomes conscious, the details of walking to the TV set and
turning the channel knob may be mostly automatic. It can be done
in a preoccupied state, for example. The conscious attentional
goal image may be broadcast to many processors, for instance, eye
movement nuclei in the brain stem. In the case of the program on
television, once the goal of watching television becomes
conscious, with no competing goal image or intention, our eyes
and head will swivel in the right direction automatically. Thus
attentional specialists may simply reside among the other
unconscious specialists, and may be triggered into action just
like the action control systems discussed in Chapter 7.
‹j ‹å 8.12 Automatic attention: unconscious control of access to
consciousness.
We already have proposed a simple mechanism whereby
different unconscious events may access consciousness: namely,
competition between input processors (2.xx), guided by feedback
From receiving processors (5.xx). However, this kind
competition is not guided by system goals. Access that is not
sensitive to goals may not be harmful when there is no urgency
and enough time to allow different elements to come consciousness
to be evaluated for relevance. But random, impulsive access to
consciousness becomes maladaptive when quick decisions are needed
to survive or to gain some advantage. We would not want a stray
thought to enter consciousness just when walking along the very
edge of a cliff, or when making split©second decisions to avoid a
traffic accident. We need some way in the GW model to connect
automatic access control to the goal hierarchy; but we cannot
afford to let existing goals control all input automatically,
because some information whose significance is not yet known may
become very important in the future. As usual, we face a trade™off
between rapid, routine access, and flexible but slower access.
The role of significance is not the only thing to be
explained; another major factor is practice. There is a sizable
research literature on attentional automatization with practice
in perceptual search (e.g., Neisser, 1967; La Berge, 1974;
Shiffrin & Schneider, 1977). We know, for example, that scanning
a list for a well-practiced word will cause the target to "pop
out" of the field. Evidently there are detection systems that
present the target word to consciousness quite automatically,
once they find it. Similarly, items that come to consciousness in
automatic memory search have a compelling, unavoidable quality,
suggesting that here too, access control to consciousness has
become automatic (Sternberg, 1963; see 2.xx). The same may be
said for the well-known Stroop phenomenon, in which the printed
înameï of a color word like "brown" or "red" tends to drive out the
act of naming the îcolorï of the word. In all these cases, access
to consciousness has itself become automatic and involuntary, at
least in part due to practice. We have previously developed
arguments that automaticity can be quite rigid and inflexible
(x.xx), and this suggests that automatic control of attention,
too, can be rigid and dysfunctional in new situations.
A simple word©game can make the point. We can ask
someone to repeat "poke, poke, poke, ..." ten times, and then ask, "what do
you call the white of an egg?" Most people (even those who really
know better) will answer "the yolk" ©©© presumably the first word
that comes to mind. Or we can ask a person to repeat "flop, flop,
flop, ..." and ask,"What do you do at a green traffic light?"
Again, most people tend to answer, "Stop" ©©© which is not
correct. Even practicing the priming word five or ten times will‹j ‹
set the system to bring rhymes to consciousness, as if conscious
access control is at least momentarily controlled by rigid
automatisms (Reason, in Baars, in press, d). Relying on automatic
conscious availability is known to lead to errors in reasoning as
well (Tversky & Kahneman, 1973).
Automatic attentional mechanisms must surely influence
spontaneous thoughts and memories outside of the experimental
setting as well. By comparison to the great research literature
on voluntary recall there is only a small body of work on
spontaneous memory, though in the real world, spontaneous
thoughts and memories are surely many times more common than
deliberate acts of recall (Klinger, 1971; Pope & Singer, 1978;
Horowitz, 1975 ab). These studies of thought monitoring indicate
that the spontaneous flow of thought is highly sensitive to
current personal significance, just as we would expect from the
evidence discussed above.
8.2 Modeling voluntary and automatic access control.
8.21 Model 6A: Automatic attention is guided by goals.
Figure 8.21 shows how automatic attention might work in a
GW framework. Suppose that there are "attentional contexts" ©©©
goal contexts whose main purpose is to bring certain material to
mind: to move the eyes and head, to search for words in memory
and bring them to consciousness, etc. These are symbolized in the
diagram by context frames with the words "get A" or "get B." We
have already discussed how the intention to retrieve a word may
operate like this (6.xx), so we are now only generalizing this
notion. In the Figure, two attentional contexts compete to guide
mechanisms for bringing different materials to consciousness. In
the beginning, context A dominates the stream of consciousness,
as it might in listening to a football game on the radio.
Stimulus B is able to interrupt the flow, by virtue of its
personal significance: it might be the listener's own name. How
can this stimulus interrupt context A? The key idea is that the
name is detected unconsciously (e.g. MacKay, 1973), and that it
activates a high level in the dominant context hierarchy in a way
that is inconsistent with current lower levels of the hierarchy.
In Figure 8.21, the listener's own name may be recognized as
important by high©level goals, which serve to disrupt the
dominant context hierarchy that controls the experience of the
football game. The hierarchy is then reconstructed so as to take
account of the new input B in the same way that a goal hierarchy
is supposed to reorganize after any surprising event (4.41). The
change in context may happen quickly enough so that the
interrupting name can be identified and become conscious.‹j ‹å
v * v©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©É
v * vInsert Figure 8.21 about here.Ƀ.<<ƒÑ
v * v©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©É
The ability to explain interruption by significant stimuli
is very important theoretically. In Chapter 1 we noted that the
various "filter" models of attention have not successfully
explained either the interruption by personal names, or the
biasing of ambiguous conscious words by unconscious material
(1.xx). These models give rise to a "filter paradox," as follows.
Let us suppose, along with conventional models, that attention
involves a selective filter that eliminates certain input from
consciousness. The filter must process and represent the meaning
of an interrupting stimulus in order to keep it from
consciousness. But if that is true, then having such a filter
would save no perceptual processing capacity, because it takes as
much capacity to detect something to eliminate it, as it does to
detect it in order to make it conscious (Footnote 1). But the
entire rationale of selective attention by filtering rests on the
assumption that it does save processing effort (Broadbent, 1958).
Hence the filter paradox.
We have previously suggested that all input, conscious and
unconscious, may be processed automatically to identify it; but
only conscious material is broadcast systemwide. If that is true,
then attention does save processing effort, but not on the input
side. If all input, conscious and unconscious, is processed
enough to identify it, the savings occur only in the fact that
unconscious material is not broadcast system©wide, and therefore
does not engage most of one's processing capacity.
We now suggest that unconscious material may either disrupt
or shape the conscious stream, if the unconscious input interacts
with the dominant context hierarchy that currently controls the
attended stream. îDisruption ïîïof the dominant context can occur if
the unconscious input activates high levels of the goal hierarchy
that are incompatible with current lower levels. If we are
listening to a boring football game and someone calls our name,
that may disrupt the dominant context hierarchy controlling the
conscious stream, because our own name is more significant than
the boring game. Alternatively, the unconscious input may remain
unconscious but help to îshapeùïú the conscious experience (e.g.
MacKay, 1973 etc.). If the unconscious input is compatible with
the dominant context hierarchy, as in the case of disambiguating
unconscious words discussed in section 1.xx, unconscious input
may help to shape the conscious experience. îïIn this way, the
unconscious word "river" can bias the conscious interpretation of
an ambiguous word like "bank" ©©© even though the unconscious
word never becomes conscious. The GW approach therefore suggests
a way of resolving the filter paradox for both of these
experimentally demonstrated cases.
The main point, again, is that automatic attention is‹j ‹
evidently sensitive to the dominant context hierarchy, and
particularly to dominant goals. Input that triggers a high©level
goal seems to receive higher access priority to consciousness.
8.22 The Options Context for conscious metacognition.
Thus far it seems that we can understand automatic attention
without adding fundamentally new ideas. Voluntary attention will
be somewhat more demanding.
In a computer one can find out the contents of any memory by
calling up a "directory," a list of files that may be selected.
Given the directory, one can choose one or another file of
information, and engage in reasoning about the different options:
one file may lack the desired information; another may have it in
an awkward format; a third may be so long that the information
will be hard to find, and so on. Seeing the directory permits one
to engage in conscious reasoning about the options. But once a
file is selected, its contents will dominate the viewing screen.
The exact same necessity and solution would seem useful in the
nervous system (Footnote 2). It would be nice to have rapid
conscious access to the alternatives which one could pay
attention to.
Even a well©known phenomenon such as Short Term Memory (STM)
suggests that there is a such a "directory" or "menu" of readily
available options. We can retrieve information from STM; we can
rehearse it, using voluntary inner speech; or we can report its
contents. People can do any or all of these separable functions
on request. Clearly all them must be readily available. But how
do we represent this fact in GW terms? We could say that certain
specialists or contexts are highly active, ready to compete for
access to consciousness. However, it seems more convenient to
represent the options in a goal context of their own, much like
the context of alternatives shown in Model 3 (Figure 5.xx).
Voluntary control of attention then comes down to ready conscious
access to, and the ability to select, the choices defined within
such an Options Context. Specialized processors could "vote" for
the various options, as shown in Figure 7.xx, and the winning
option could evoke the appropriate effectors and subgoals by
ideomotor control, as in Models 4 and 5 (6.xx and 7.xx). Thus we
seem to have all the makings for a mental directory already at
hand, as shown in Figure 8.22.
v * v©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©É
v * vInsert Figure 8.22 about here.É
v * v©©©©©©©©©©©©©©©©©©©©©©©©©©©©©©É
‹j ‹å An Options Context shows consciously available options for
future conscious contents. It is shown in Figure 8.22 as a
rounded frame with the options listed inside. Options Contexts
may be evoked from among the receiving processors in the GW
system by a conscious event, such as the question, "What should I
do next?" Once the Options Context dominates the global workspace
it presents a "menu" or "directory" of possible conscious
contents; the most relevant one is chosen by a decision process,
which may include "votes" from specialists, as well as from the
Goal Hierarchy; and the winning option then evokes a working
context by ideomotor control. This may seem complex, but it is
difficult to see a simpler way to implement what we know about
voluntary attention.
Figure 8.22 shows how an Options Context might work in the
case of Short Term Memory (STM). The simplest event that may
evoke STM might be a conscious query about some recent event to
be recalled ©©© "What just happened?" ©©© but Short Term Memory
is obviously used îin the service ofï many other functions. For
example, if we have a goal of gaining familiarity with a certain
psychological theory, STM might be used to connect two recent
ideas which were not previously connected. In that case the
conscious query that evokes the STM options context might be,
"How does the idea of 'attention' relate to 'consciousness?'"
Options contexts may be used as subgoals in the service of any
dominant goal, including another options context. Thus the
Self™monitoring Options Context in Figure 8.22 presumably makes
routine use of Short Term Memory.
8.23 Model 6B: Voluntary attention allows conscious choice.
We have previously noted that voluntary attention must be
sensitive to the goal hierarchy, but that it cannot be completely
controlled by automatic goals. After all, we can consciously work
to change our own goals (8.24), and even the most habit©driven
person can for some time avoid paying attention to habitual
conscious contents. This may be quite difficult at times, but if
it were impossible, people would lose the ability to change their
goals. Flexibility is indeed the only reason to have voluntary
attention in the first place. Otherwise automatic attentional
control would be quite sufficient. So somehow the current model
must reflect the evidence that voluntary attention retains some
amount of freedom from automatic control by the goal hierarchy.
Inherent in the global workspace, of course, is the notion
that multiple sources of information can interact to create new
responses to new conditions. Figure 8.22 shows how a coalition
of specialized systems can "vote" for one or another option,‹j ‹
something that is not possible with automatic attentional
contexts as shown in Figure 8.21. Nevertheless, goal contexts can
still influence the choice, and indeed, different parts of the
goal hierarchy may support different conscious contents. From
this point of view, conscious choices may cause the goal
hierarchy to decompose; previously cooperating goal contexts may
now compete in support of different options. This is indeed the
definition of a dilemma ©©© being caught between one deep goal
and another. This is of course the stuff of human tragedy, on
stage and off.
Conscious inner speech may be very important in maintaining
voluntary attentional control when it runs counter to automatic
tendencies. We have already mentioned the evidence from clinical
work with children who have impulse©control problems, showing
that teaching them to speak to themselves is helpful in
maintaining control (Meichenbaum & Goodman, 1971). On a short™term basis,
voluntary inner speech, and voluntarily retrieved
mental images, may also be helpful to fight off automatic
attentional tendencies. In the model, conscious availability of
recent thoughts ©©© "Do your homework, don't think about playing
outside!" ©©© may help to control the automatic tendencies, at
least for a while. Thus recency may be used to combat
automaticity. But a permanent victory for the voluntary control
presumably requires the creation of a new, coherent context,
within which the automatic choices are differently defined
(4.xx).
v * v-------------------------------É
v * vInsert Figure 8.23 about here.É
v * v-------------------------------É
8.24 Automatizing voluntary attention.
Experimental studies of trained perceptual access may mimic
the way in which normal attentional mechanisms become automatic.
In a typical experiment, subjects are asked to pay attention
voluntarily to something they would normally ignore (e.g.,
Neisser, 1967). Subjects come into the experiment of course with
their own goals, from earning money to impressing the
experimenter. In order to achieve these goals, they are asked to
do something that was previously quite irrelevant to them.
Searching for a conscious stimulus ©©© a famous face in a picture
of a crowd, for example ©©© is given very high priority by the
experimental instructions. The instructions say, in effect, that
in order for the subject to perform satisfactorily, he or she
must pay attention to the face in the crowd. The task is repeated
over and over again until it becomes automatic ©©© that is, until‹j ‹
the alternatives in the voluntary options contexts for the task
are reduced to one (5.xx).
If we can generalize from this situation, there are
apparently two necessary conditions for creating automatic access
to consciousness:
1. A target that has low priority for access is given high
priority by a temporary, consciously available goal
context, which may be associated with high
levels of the permanent goal hierarchy. E.g. social
compliance in an experiment.
2. Voluntary attention to the target is practiced until the
options context has no more degrees of freedom, so
that it changes into a single goal context (Fig. 8.xx).
Presumably one's own name acquires associations early in
life with high priority goals, such as one's desire for
attention, for protection and care, for food, for avoiding
punishment. And surely paying voluntary attention to one's name
occurs many thousands of times. Hence, presumably, the Moray
phenomenon of the subject's name breaking through to
consciousness by virtue of automatic access control.
We can now consider two important cases of attentional
control, directing attention toward something versus directing it
away from something. The first is obviously important, and the
second raises the classical psychodynamic issues of suppression
and repression, which can be easily modeled in GW theory.
8.3 Directing attention îtowardï something.
Suppose we feel hungry, and have some conscious image of
delicious food. Inherently, this image, we have argued, recruits
processors able to help achieve the goal, and these must include
attentional processors. If we are able to reach for food
automatically, little attentional control would seem to be
required. But if we must think about how to reach the desired
food, to deal with obstacles, or to make choices about equallly
attractive alternatives, the goal image should be able to recruit
access of these issues to consciousness. Thus, given an
interesting conscious goal image, recruitment of attention should
happen automatically along with recruitment of other subgoals to
the goal. Thus the simplest case of directed attention toward
something involves one goal image recruiting automatic eye‹j
‹´movements, attentional search, etc., in order to bring up a
conscious content.
8.31 Using voluntary attention to create new access
priorities.
One way to make an unimportant stimulus important is to
associate it explicitly with one's major goals. This is indeed
what one does in conditioning. Pavlov's dog is typically deprived
of food for a day or so, so that eating is high in the goal
hierarchy. Through paired repetition, the famous bell then
becomes a signal for food, so that it functions as a conscious
event that engages the eating goal. Similarly, in operant
conditioning, the act of wiggling one's tail may become a
subgoal, after which, magically, food appears in the Skinner Box.
While one must be cautious in anthropomorphizing animal
experience, surely the experience of humans in such a situation
is easy to guess: it always involves explicit, conscious
association of the conditioned event with a pre©existing,
significant goal. "Aha! so pushing this button always gives the
right answer."
Nor is conscious association of new events with existing
high©level goals limited to the laboratory. One extremely common
persuasive technique used by all politicians and advertisers is
to associate a previously irrelevant event with a major life goal
of the audience. Underarm deodorant was not very important before
millions of people were consciously reminded that it is a îsine
qua nonï of social acceptability. For our purpose, this suggests
two critical points: one, that the event to be connected to the
significant goal must be conscious; and two, that this event can
then come to control attentional mechanisms that control access
of the previously irrelevant event to consciousness.
8.32 Mental effort due to competition between voluntary
and automatic attention.
We have defined "mental effort" as voluntary control against
unexpected resistance (7.xx). One obvious example involves trying
to control one's attention voluntarily against contrary automatic
tendencies.
Every child knows how it feels to have homework when s/he
really wants to go outside and play. The process here presumably
involves decision©making (7.xx), except that the decision is not‹j ‹
in the first instance about îdoingï something --- it is about
îpaying attention toï something. That is, it is a struggle whose
first outcome is purely mental. The act of paying attention to
homework is more novel and effortful and less pleasant, and hence
requires more conscious involvement, than the routine and
pleasant act of thinking about playing. But once the issue is
decided, one may become absorbed in the chosen path --- which is
itself controlled by unconscious contexts, naturally. Then the
experience of struggle and effort may utterly disappear, even for
tasks that were initially seen as onerous and boring. Absorption
is typically experienced as positive, perhaps because we simply
do not monitor our experiences when we are truly absorbed (8.xx).
Or perhaps absorption is inherently pleasant as well
(Czikzsentmihalyi, 19xx).
This kind of mental effort is a struggle between voluntary
and automatic attention. Presumably the conscious goal images of
having to do homework serve to recruit additional processors and
contexts which may be able to "outvote" the automatic attentional
systems promising fun and adventure by playing outside. This
voluntary decision may have to be repeated a number of times at
natural choice-points in the task. After taking a break from
doing homework, the same struggle may have to be repeated. At
each decision point, the intention to think about playing outside
will tend to come up automatically, while the intention to
continue with homework will have to be raised voluntarily,
perhaps with the aid of a powerful motivating goal image or inner
speech reminder. Each decision point can be viewed as a case of
the "voting process" discussed in section 7.xx, but one that is
îmetaïcognitive --- i.e., it concerns what is to become conscious
later.
Obviously some of the recruited voting systems may be more
significant than others. If some deep goal context is recruited
to support homework --- such as the promise of quick approval by
a loving parent, or the threat of ridicule by an older sibling
--- the relevant deeper goal contexts can move the vote in one
direction or another (Figure 8.1x). During the decision struggle,
conscious thoughts related to these deeper goals may aid one or
the other side in the process.
We can see this struggle for control especially in vigilance
tasks, in which people are asked to monitor a blip on a radar
screen, or some other minor perceptual event in an otherwise
boring situation. Attentional control drops quite rapidly under
these circumstances (Mackworth, 1970). But variation in
attentional focus is quite normal even in interesting tasks
(Warm, 1984). In all of these cases, one may enter a period of
conflict between voluntary tendencies to continue attending with
high efficiency, and the spontaneous or automatically controlled
tendency to attend less well.
‹j ‹å
8.33 The importance of accurate source attribution.
For effective metacognitive control, we should be able to
refer to previous or future events with accuracy. If we decide to
repeat something we just learned, by the arguments of Chapter 7
we should be able to retrieve the relevant goal image and guiding
goal context. Similarly, if we are to answer questions about our
own thoughts, we must be able to refer to them. îAccurate source
attributionï ©©© knowing why and how we did what we did ©©© seems
vital for these tasks. And yet, we know that source attribution
fails in a number of important cases (Nisbett & Wilson, 1977;
Ericsson & Simon, 1984; Langer & Imber, 1979; Weiner, 19xx). That
is, much of the time we lose track of why and how we did what we
did, so that it becomes difficult to report these things
accurately.
We have discussed the Langer & Imber (1979) study, in which
automaticity in a task undermined the ability of people to report
the features of the task, and made them more vulnerable to a
misleading attribution about the task. In terms of the model, we
can simply see how a voluntary options context is reduced to a
single option (that is, just a single goal context). Further,
goal images should become more fleeting over time if they are
predictable. In either case, it should become much more difficult
to retrieve the way in which one performed the task.
8.4 Directing attention îaway fromï something: Suppression,
repression, and emotional conflict.
Although there is unresolved controversy in the scientific
literature about the existence of repression, there is no real
doubt about the existence of some sort of tendentious evasion of
conscious contents. Even the most skeptical observers acknowledge
that people tend to make self©serving errors and interpretations
whenever there is enough ambiguity to permit this. The scientific
arguments seem to revolve around the issue of repression, defined
as îunconsciousï inhibition of troubling conscious contents (e.g.
Holmes, 1972, 1974; Erderlyi, 1985). This is of course a
terribly contentious issue, because we do not currently have good
scientific tools for assessing these "zero©point" issues (1.xx).
It is very difficult to tell with certainty whether someone
really knew, even momentarily, that a thought was sufficiently
threatening to avoid. But the fact of avoidance, and even some of
the mechanisms of avoidance, are not really in question.
‹j ‹å How can we possibly avoid thinking of some topic if the
decision to avoid it is itself conscious? Children play a game in
which the participants try to avoid thinking about pink
elephants, and of course they cannot do so. There is a
contradiction between having the guiding goal image "pink
elephants" and the trying to avoid it at the same time.
Similarly, in clinical hypnosis there is extensive lore
suggesting that subjects should not be given negative
suggestions. To help someone to stop smoking it is not helpful to
say, "You will stop thinking of cigarettes," because the
suggestion itself contains the thought of cigarettes. Thus
directing attention îaway from ïsomething seems to be quite
different from directing it îtowardï something.
8.41 Thought avoidance can occur in many ways.
The "pink elephant" game cited above should not be taken to
suggest that people simply cannot avoid conscious thoughts in a
purposeful way. There is very good evidence for the effectiveness
of thought avoidance, both in normal and clinical experiments
(e.g. Meichenbaum & Bowers, 198x). Several mechanisms may serve
to exclude information from consciousness. These range from
changes in receptor orientation, as in evasive eye©movements, to
deliberate failure to rehearse items in short©term memory in
order to forget them, or tendentious re-interpretation of
experiences and memories (Bjork, 1972; Loftus & Palmer, 1974;
Holmes, 1972).
"Directed forgetting" is one example (Bjork, 1972). People
in a short©term memory experiment are simply told to forget
certain items, and they will do so quite well. They probably
rehearse only what they need to remember, and the to©be™forgotten
items fall by the wayside. That is, one can use the
limited capacity of short©term memory to load one topic in order
to avoid another. This is indeed the principle of distraction:
every human being must engage in some distraction sometimes to
avoid pain, or boredom, or difficulty. In these cases there is no
doubt about the existence of conscious thought avoidance.
In general, we can distinguish between îstructuralï and
îmomentaryï evasions. A religious belief system may help the
believer escape anxiety about death and disease. After all,
almost all religions provide reassurance about these things. Once
a reassuring belief system is accepted and not challenged it‹j ‹
creates a conceptual context for the believers' experience
(4.xx); thus certain anxious thoughts presumably will come to
mind much less often. There is a close connection between this
common observation and the claim made in Chapter 7 that the
ideomotor theory suggests an interpretation trust and confidence
in terms of a low degree of competition between different
conscious thoughts (7.xx). Suppose that thought A is anxiety
about disease, and not©A proclaims that disease is only a trial
on the way to heaven; if not©A, once thought, is not contradicted
then there is a conclusive answer to the source of anxiety.
Probably most human beings in the world operate within
self-serving belief systems of this kind. Their prevalence
suggests that reassuring belief systems are quite effective,
often for a period of years.
There are obviously also îmomentaryï evasions, such as not
looking at beggars on the street, avoiding the gaze of dominant
or frightening persons, and avoiding recall cues for unwanted
memories. In principle any of these mechanisms can come under
purposeful control, either voluntary or involuntary.
8.42 When we lose voluntary access to avoided thoughts,
there is the appearance of repression.
In fact, the clinical evidence for repression is just
îapparently purposeful but disavowed failure in voluntary access
to normally conscious eventsï. If we fail to recall some painful
event that happened just yesterday, even though we remember
everything else; or if we cannot remember a thought that
challenged a fundamental belief; or if we fail to make an obvious
inference that seems obvious but is or painful to contemplate;
any cases like this are likely to be interpreted as repression
clinically. One classical example is the "belle indifference" of
the conversion hysteric, who may be suffering from psychogenic
blindness, local anesthesia, or paralysis, but who may deny that
this is much of a problem (Spitzer, 1979). Thus the key is
failure of voluntary, metacognitive access. (Involuntary measures
of memory may not show any decrement: we know that recognition,
skill learning, and savings in recall may survive failures of
voluntary recall (Bower, 1986).) But of course we can model such
access failure from Model 6 (8.xx). Figure 8.42 shows how the
outward evidence for repression may be modeled in GW theory.
-------------------------------------------------------------
Insert Figure 8.42 about here.É
-------------------------------------------------------------
8.43 Signal anxiety as ideomotor control of thought avoidance.
There is a very interesting connection between these ideas
and the psychodynamic notion of "anxiety as a signal." GW theory
suggest that people may have fleeting quasi-conscious goal images
that may serve to mobilize attentional mechanisms for avoidance
of certain conscious contents. This is precisely the role of
signal anxiety. While the notion of signal anxiety may sometimes
apply to clearly conscious feelings, some sources suggest that
people can have very fleeting images that serve as warnings to
avoid certain upsetting thoughts. Thus Freud is quoted as writing
that thinking must aim "at restricting the development of affect
in thought-activity to the minimum required for acting as a
signal" (Freud, 1915/1959). In discussing the appearance of
"substitutive ideas" in phobia, i.e. ideas that may evoke less
fear than the original phobic object, Freud writes that
"Excitation ... give(s) rise to a slight development of anxiety;
and this is now used as a signal to inhibit ... the further
progress of the development of anxiety." Here is another point of
theoretical contact. Note that one the same ideas play a role in
behavior-modification theory of phobia. For example, one can have
a hierarchy of increasingly upsetting mental images about fire. A
fire phobic may be able to have the thought of a book of matches
with little fear, but the thought of a bonfire may be quite
frightening. The image of a matchbook may then act as a safe goal
image, which may trigger avoidance mechanisms that help the
person to stay away from the really troubling mental images.
8.5 Further implications.
8.51 Absorption and suspension of disbelief.
Conscious metacognition should of course compete with other
limited-capacity events, including the conscious content that is
being controlled attentionally. We cannot read this book and at
the same time entertain the conscious possibility of doing other
things. It follows that absorption in a stream of events, such as‹j ‹
a movie or piece of music, should decrease access to
metacognitive options. One common observation is that when we
become absorbed in a film or novel, we can easily identify with
the main characters. In the parlance of the theater, we "suspend
disbelief". If disbelief is a conscious disputing of previous
conscious contents (7.x), this is easy to model in GW theory. We
need only suppose that disbelief requires an options context of
the form, "Is what I just experienced really true or acceptable?"
In deeply absorbed states accessing these conscious options may
be competed out of consciousness. Suspension of disbelief then
presumably liberates our tendencies to identify with attractive
fictional models, free from inhibition and contradiction. We can
allow ourselves for a while to live in wishful fantasy.
8.52 Hypnosis may decrease access to options contexts.
Chapter Seven suggested that hypnosis is reducible to
absorbed ideomotor control. If that is true, and if absorption
implies a decrease of access to attentional options, then we may
be able to explain the extraordinary compliance of hypnotic
subjects. We may suggest that self©other differentiation often
requires a momentary conscious decision: "Did I really want that,
or was I persuaded by advertising to want that?" "Did the
hypnotist tell me to raise my arm, or did I tell myself to do
so?" Compliance in high©hypnotizable subjects may therefore
follow from their capacity to be deeply absorbed in the hypnotic
situation, to the point where no conscious capacity is available
to reflect on the situation from an outside perspective.
Previously we were able to account for several other features of
hypnosis (7.xx), but not the remarkable compliance with
suggestion, and the lack of resistance of unusual suggestion. We
can now fill in this gap.
8.53 The role of attention in motivation and the maintenance
of mental stability.
There are thousands of experiments focused on "perceptual
defense," the apparent tendency of people to avoid reporting
rapidly flashed words that are obscene or conflictful (Erdelyi,
1974). However, these experiments apparently showed two opposite
tendencies. People sometimes underreported taboo words
("perceptual defense") and sometimes overreported them compared
to control words ("perceptual sensitization"). This seemed to be
a paradox, which led to considerable criticism and
disillusionment. However, the coexistence of "defense" and‹j ‹
"sensitization" may not be just an experimental difficulty, but a
fundamental fact about human allocation of attention. After all,
we must do two things when presented with something painful,
alarming, or ego©threatening: first, we must know that it is
there, so that we can cope with it; second, if possible we try to
avoid it. If the event is novel, we presumably need conscious
involvement in order to identify it and to learn to avoid it.
Thus the existence of both sensitization and avoidance is
something one might predict on an a priori basis.
This suggests that attention may have two major functions:
(1) Allocation of conscious involvement to significant
events and problems, by making them conscious in proportion to
their motivational significance; this includes painful, alarming,
or ego©threatening events. Some examples include the resistance
of significant stimuli to habituation; the ability of one's own
name to interrupt ongoing streams of conscious events; and our
tendency to pay more attention to problems that demand more novel
solutions.
This role of attention may be countered by another function,
which is,
(2) Regulating the flow of novel information, so that we do
not confront either too much, or the wrong kind of novelty. We
can think of this as protecting the context hierarchy from
too-rapid change. Thoughts that threaten the stability of one's
beliefs are avoided by most people.
Notice that in the case of painful sources of information,
these two tendencies will run counter to each other: on the one
hand, pain or threat is important, and therefore deserves
attention; on the other, it is may demand such a fundamental
change in the context hierarchy that it is rather avoided.
It is possible that psychodynamic though©avoidance result
From this second role of attention. Excessive novelty, especial
emotionally charged novelty, may threaten a fundamental
realignment of the goal hierarchy. We connect this hierarchy to
the notion of "self" in the next chapter (9.0).
8.54 Some further thoughts about the operational definition
of conscious experience.
Chapter One suggested an operational definition for
conscious experience, one that we have tried to adhere to quite
faithfully. Namely, we were going to consider something a
conscious experience if it could be reported accurately, and was‹j ‹
claimed by the subject to be conscious. Early in the theoretical
development we could not model this phenomenon of accurate
retrospective report, of course, because it is actually quite
complicated. It involves
--- consciousness of an event (e.g. 5.xx)
--- retrospective ability to direct conscious recall
voluntarily to the event (e.g. 8.xx)
--- the ability to recode the event in words or gestures
(e.g., 7.xx)
--- the ability voluntarily to carry out those words or
gestures. (e.g. 7.xx).
In brief, in order to model the operational definition we
first needed a usable conception of consciousness, volition, and
attentional access. We had to "bootstrap" upward from the
operational definition, until eventually we could hazard a true
theoretical account. Ultimately, of course, every theory must
explain its own operational definitions. Are we ready to do so
now?
Notice that initially, there is merely a conscious
experience of a banana. Numerous systems adapt to this conscious
event, including systems able to re-present a later image of the
conscious event, voluntarily, on cue, in the context provided by
a goal system that is instantiated by the experimental
instructions. This goal system, when it is triggered by the
question, "What did you just experience?" acts to guide a search
for a plausible answer. Norman & xxxx have pointed out that
finding a plausible answer to any question, such as "What was
George Washington's telephone number" requires a complex and
sophisticated search, one that we would framew within a goal
context. Similarly, a subject who has been instructed to look at
a television screen and perceives a banana, must know that the
correct answer is "banana" and not "television screen". So there
is a large interpretive component in answering this question.
Once it is interpreted properly, it is reasonable to think
that one can voluntarily attempt to recall recent events, decide
which ones could be meant by the questioner, retrieve it as an
image, redisplay it consciously, and allow unconscious verbal
systems to search for a lexical match: "a banana." Again, it
takes a voluntary act to carry out the verbal report, which
involves by the arguments of Chapter 7 a momentary conscious goal
image of the distinctive aspects of the action (perhaps the word
"banana" in inner speech), the goal image is rapidly inspected by
numerous systems able to check its propriety, and barring
contrary images or intentions, it executes. Complex? Certainly,
yes. But probably not too complex in the face of psychological
reality.
It is quite satisfying to be able to give a plausible
descriptive account of one's operational definition, after seven‹j ‹
or eight chapters of hard work. It suggests again that we are on
the right track.
8.6 Summary.
This chapter has explored the topic of attention, defined as
access control to consciousness. There is a traditional
distinction between voluntary and involuntary attention of which
we can make good use. Automatic attention apparently involves
unconscious mechanisms for bringing things to consciousness, in a
way that is controlled by the goal hierarchy, so that significant
things have access priority. Voluntary attention requires a new
idea: the notion of an Options Context, which presents
alternative things one could aim to make conscious next. As
before, these comparatively simple ideas have widespread
implications for other psychological questions. There is a
distinction between calling attention to some topic, and steering
attention away from some painful or threatening topic. Both kinds
of phenomena can be accomodated in the GW framework. Finally, we
explored some of the implications for absorption, hypnosis, and
the operational definition of conscious experience that was
proposed in Chapter 1.
Footnotes
1. I am grateful to Michael Wapner for clarifying the "filter
paradox" for me.
2. I am grateful to David Spiegel and Jonathan Cohen for
suggesting the directory analogy in the context of hypnosis.