Part III. The fundamental role of context.
The next two chapters explore the role of ?contexts? ---
unconscious systems that evoke and shape conscious experience.
Chapter 4 maintains that context effects are pervasive in all
psychological domains. We survey the extensive evidence for this
claim, the various kinds of contexts, and the ways in which they
may interact. In a sense, contexts can be thought of as
information that the nervous system has ?already? adapted to; it is
the ground against which new events are defined. Consciousness
always seems to favor novel and informative messages. But
recognizing novelty requires an implicit comparison to the ?status
quo?, the old knowledge that is represented contextually. Chapter
5 develops the notion that all conscious events provide
?information? by reducing uncertainty within a stable context.
<
Chapter Four
Model 2: Unconscious contexts shape conscious experience.
4.0 Introduction.
4.1 Sources of evidence on contexts.
4.11 Priming effects:a conscious experience improves
receptivity to related conscious experiences.
4.12 Fixedness: being blind to "the obvious."
4.13 Top-down influences and the pervasiveness of
ambiguity.
4.14 Decontextualization: Strong violations of context can
become consciously accessible.
4.15 A summary of the evidence for unconscious contexts.
4.2 Several kinds of contexts.
4.21 The Context of Perception and Imagery.
4.22 The Context of Conceptual Thought.
4.23 Intentions as Goal Contexts.
4.24 Other contexts.
4.26 The different kinds of context interact.
4.3 Modeling contextual knowledge.
4.31 Contexts as stable coalitions of processors.
4.32 The current Dominant Context imposes unconscious
constraints on what becomes conscious.
4.33 Contexts do not Œcompletely‹ constrain conscious
experiences.
4.34 Internal consistency and the role of organization in
contexts.
4.35 Cooperating and competing contexts: Model 2.
4.4 Some plausible properties of contextual knowledge.
4.41 Accessing and leaving contexts.
4.42 Maintaining contexts with conscious reminders.
4.43 Modifying contexts: coping with the unexpected.
4.5 Implications for empirical testing. <j <†
4.51 Related findings.
4.52 Some testable predictions from Model 2.
4.53 Some questions that Model 2 does not answer.
4.6 Summary and a look ahead.
< <
4.0 Introduction.
Imagine stepping on a small sailboat on a fine, breezy day,
and setting off for a short sail. The weather is fair, and as you
set sail from the harbor the water becomes choppy but not
uncomfortable. At first, the horizon seems to swing up and down,
but you quickly realize that it is the boat that is moving, not
the horizon. As you gain your sea legs, the world becomes much
steadier. On the way home the movements of the boat seem almost
placid, though the force of the wind and the waves has not
changed. Your sailboat is tied up to the dock, and as you step
back on dry land, the horizon suddenly seems to sway, and you
must steady yourself; but very quickly the world becomes stable
again.
This everyday experience sums up the topic of this chapter
and the next. As we walk, run, turn, sit, dance or climb on dry
land, specialized components of the nervous system make running
predictions to compensate for our changing relationship to
gravity and to the visual surround. The world is experienced as
stable only when this remarkable feat of prediction is
successful. These orientational predictions are entirely
unconscious, but they profoundly influence our conscious
experience. As long as they are successful, these contextual
predictions give no sign of their existence. That may change for
a time when we step on a small sailboat, but in a curious way: we
still do not experience the change in the Œframework‹ of our
experience, we just notice an instability in the Œentire
perceptual field‹. Stepping on the sailboat we experience the
novel, unpredictable movements of our body as a change in the
world, even though we know full well that the world has not
changed; only our relationship to it has. The real world does not
sway with the motion of the deck. The same thing happens for a
moment when we step back on dry land after "gaining our sea
legs": unconsciously we now predict a regular yawing and rolling,
so that the relationship between reality and expectation has once
more gone awry. This experience of an unstable world causes the
contextual orientation system to revise its predictions again,
and since we are experienced land walkers, we soon regain our
feet.
The system that computes our orientation to gravity and the
visual world is part of Œthe context‹ of our experience. We
continually benefit from a host of such contextual processes,
without experiencing them as Œobjects‹ of conscious experience.
Their influence can be inferred from many sources of evidence.
The example of the sailing trip involves perceptual-motor
context, but much the same argument can be made for the contexts
of thinking, belief and communication (4.x). A great deal of the
research literature in perception and cognition provides evidence
for the pervasive influence of unconscious contexts (e.g., Rock,
1983; Bransford & Franks, 1976; Levicky, 1986). (Footnote §1) <j <†
"Context" is a key idea in this book. Chapter 2 defined
"context-sensitivity" as the way in which unconscious factors
shape our conscious experience. Chapter 5 will suggest that
habituated or automatized processes do not disappear, but become
part of a new context that will shape later conscious
experiences. A Œcontext‹ is thus Œa system that shapes conscious
experience without itself being conscious at that time.‹ §2 It is a
close modern relative of "set" and "adaptation level" in
perception (Bruner, 1957; Helson, 1964; Allport, 1954, Uznadze,
19xx), and of various proposals for knowledge structures and
"frames" in cognitive science (Minsky, 1975; Clark & Carlson,
1981). Contexts are equivalent to currently unconscious
Œexpectations‹ that shape conscious experiences, and to currently
unconscious Œintentions‹ that shape voluntary actions (see 6.0 and
7.0). The observations supporting this idea were well-known to
precbehavioristic psychologists in Europe and the United States,
including Wundt, James, the Wu“"rtzburg School, Brentano, Gestalt
psychology. and the psychologist Narziss Ach (Murray, 1983;
Blumenthal, 1977; Rapaport, 1951). There is nothing really new
here, except for a modern theoretical framework --- and the fact
that modern psychology has neglected this evidence for so long.
The word "context" is often used in current psychology to
mean the physical surround, but in this book it only refers to
the Œinner‹ world that shapes our experience. After all, the
physical environment affects our experiences and actions only if
it is represented in the inner world. Thus the context-in-the-
world inevitably shapes our experience by way of the context-in-
the-head. Further, the inner context preserves important
information from the past, which is not available from our
current surroundings at all. It makes more sense, therefore, to
locate the psychological context inside the nervous system.
Contexts are similar to "activated knowledge structures,"
"mental representations," "semantic networks," "frames,"
"schemas," "scripts," "Plans," "expectations," and other kinds of
knowledge representation that are widely discussed in the
cognitive sciences (Norman, Rumelhart, Mandler, Minsky,
Bransford, Polanyi, Piaget, Bartlett, Helson; Miller, Galanter &
Pribram ). We will borrow freely from this literature. But why
add one more term to the current rash of words that mean much the
same thing? The reason is simple. For us, the word "context" is
not just any mental representation --- it is an Œunconscious‹
representation that acts to influence another, conscious
representation. This special meaning is not captured by any of
the other terms.
This chapter will look into some of the characteristics of
stable contexts. We begin with a survey of the great amount of
evidence for contextual knowledge, specify some common properties
of contexts, and explore the interaction between conscious
contents and unconscious contexts.
<j <†
4.1 Sources of evidence on contexts.
Contexts are a bit tricky to think about, because by
definition we do not experience them directly. For this reason,
we begin with four pervasive sources of evidence for unconscious
contexts that shape conscious experience.
(1) the existence of Œpriming effects‹, where one conscious
experience alters the processing of another, although the first
experience is gone by the time the second arrives;
(2) the universal phenomenon of Œfixedness‹, where one cannot
escape the influence of unconscious contextual assumptions that
stand in the way of solving a problem, or of perceiving an
alternative;
(3) the case of top-down contextual influences, which Œchange
our conscious experience‹ of any event that is ambiguous, unknown,
degraded, fragmentary, isolated, unpredictable, or partly
forgotten;
(4) the case of Œstrong violations‹ of contextual
expectations, which can cause a part of the unconscious context
to become conscious and reportable.
Table 4.1 summarizes the contrast between conscious and
unconscious phenomena connected with context.
--------------------------------------- Table 4.1 ------------------------------------------
ŒContexts of a single conscious event.‹
ŒConscious‹ ŒUnconscious‹
1. Percepts, images, inner Contextual factors
speech, and bodily feelings that shape and evoke
these conscious events.
Immediately accessible concepts. Conceptual presuppositions.
2. Input that can be "Acontextual" input
interpreted within a (for which context
dominant context. does not yet exist,
or for which it is not
dominant, as in selective attention.)
3. Previously unattended events Unattended events that
interrupting the attended affect the
stream (e.g. the subject's interpretation of
name). attended events (e.g.
disambiguating words).
4. Strong violations of Weak violations of
unconscious contexts. unconscious contexts
(decontextualization) (e.g. proofreader effect).
----------------------------------------------------------------------------------------
We will give examples of each case.
4.11 Priming effects: conscious experiences generally
improve receptivity to related conscious experiences.
------------------------------
INSERT FIGURE 4.11 ABOUT HERE.
------------------------------
When one experience affects the likelihood of a similar
experience, we can say that the first event has "primed" or<j <
shaped the context for the second event. This is a phenomenon of
extreme generality. Blumenthal (1977) quotes Fraisse (1963, p.
88), for example:
"When I listen to speech, I perceive the clause being pronounced
by the speaker, but I interpret it in accordance with all the
sentences which I no longer perceive and of which I have only
retained a general idea. When I listen to music, I perceive again
and again a short, rhythmic structure, but this is integrated
with a melodic whole to which its owes its affective resonance."
Music and speech are indeed very good examples.
Psycholinguistic research has now amassed extensive evidence for
widespread discourse relations that are necessary to understand
even a single word in a conversation, although those
relationships are of course not focally conscious (Clark & Clark,
1977; Clark & Carlson, 1981; see 4.13). Similarly, in a piece of
music the key, the initial statement of the themes, their
development and variations, all must shape the way we experience
a single phrase in the middle of a symphony. But none of that is
conscious when we have that experience.
Figure 4.11 gives an example of a short-term priming effect.
The middle Figure (B) is an ambiguous Necker cube. By paying
attention to Figure A for several seconds and then going back to
B, we tend to interpret the ambiguous figure as if we are looking
at the bottom of the cube. The experience of Figure A structures
the experience of B, even though A is not conscious when it does
so. Now we can pay attention to Figure C for a while; now we are
more likely to see the ambiguous cube from the top.
We can easily show linguistic priming. Compare the next two
examples, in which the first word primes an interpretion of the
second:
"volume: book"
Œversus‹
"arrest: book".
The conscious interpretation of "book" will differ depending
upon the prime.
In general, a conscious priming event:
(1) decreases reaction time to similar conscious events;
(2) lowers the threshold for related material that is near
the perceptual threshold, or is ambiguous, vague, fleeting,
degraded, badly understood, or isolated from its surround. Indeed
any task that has an unconscious Œchoice-point‹ in the flow of
processing is sensitive to priming for the relevant alternative
(Baars, 1985). <j <†
(3) a prime increases the likelihood of similar events
emerging in memory, through free association, cued recall, and
recognition tasks.
(4) finally, a conscious prime stimulates the probability of
actions and speech related to the priming stimulus (Baars, 1985;
7.4).
Priming effects are ubiquitous in sensation, perception,
comprehension, and action. In an older psychological vocabulary,
priming creates set (Œ‹Œ‹Luchins, 1942; Bruner, 1957; Ach, 1905).
Indeed, the Psychophysical Law, the oldest and one of the
best-established findings in psychology, states that the
experienced intensity of any stimulus depends on the intensity of
preceding stimuli. This can be thought of as a temporal priming
effect with universal application.
Priming effects are not always momentary; they can last as
least as long as a conversation (Foss, 1982), and we suggest
below that some contexts triggered by conscious experiences last
for years (x.xx). So we are not talking merely of momentary
events. Even a single conscious experience may trigger a short
term change in context (Levicky, 1986); in the case of
traumatic experiences the effects can last for years (Horowitz,
1975 ab).
Generally the word "priming" is used to refer to those cases
where our perception of an event Œis improved by‹ earlier similar
experiences. There are also cases of Œcontrast‹, where an earlier
event causes the later one to be perceived in an opposite way.
For example, Uznadze (19xx) showed that sometimes the perceived
weight of a rubber ball will be Œincreased‹, and sometimes it will
be Œdecreased‹ by earlier exposure to another weighted ball. All
these effects can be treated as context effects by our
definition, since the observer is not conscious of the influence
of the earlier event Œat the time‹ the later one becomes conscious.
But we will focus on the more common case, where later processes
are facilitated by an earlier conscious experience.
The similarity of the prime and the primed event can be
either perceptual or conceptual. The similarity between "book"
and "volume" in the example above is not perceptual but semantic
or conceptual; the similarity between the two views of the Necker
cube in Figure 4.11 is more perceptual.
The predictions made about contexts throughout this book can
often be tested with priming tasks. This is one reason to
emphasize the role of priming.
<j <†
4.12 Fixedness: being blind to "the obvious".
The four sentences below are normal, coherent English
sentences:
(1) The ship sailed past the harbor sank.
(2) The building blocks the sun shining on the house faded
are red.
(3) The granite rocks by the seashore with the waves.
(4) The cotton clothing is made of grows in Alabama.
(Milne, 1982).
On first reading these sentences, most of us feel "stuck";
they do not cohere, they do not work somehow. We may be driven to
try rather farfetched ideas to make sense of them: maybe sentence
(1) is really two conjoined clauses, such as "The ship sailed
past Œand‹ the harbor sank"? But harbors do not sink, so that
interpretation does not work either. If we truly believe that
these are normal English sentences, the experience of trying to
understand them can be intensely frustrating and annoying.
What is going on? Consider the following Œcontext‹ for
sentence (1):
"A small part of Napoleon's fleet tried to run the English
blockade at the entrance to the harbor. Two ships, a sloop and a
frigate, ran straight for the harbor while a third ship tried to
sail Œpast‹ the harbor in order to draw enemy fire. The ship sailed
past the harbor sank."
If you have just encountered sentence (1) for the first
time, this little story should help solve the problem. Oh! You
mean "The ship (comma) sailed past the harbor (comma) Œsank‹!" But
that's dirty pool! Not so; the sentence is really quite normal,
as we can see when it is put in context.
We could, of course, insert the subordinate clause marker
"which" to create:
(1') The ship Œwhich‹ sailed past the harbor sank.
But this use of "which" is optional in English, though we
tend to insert it when needed for clarity.
The problem we encountered with sentence (1) is one kind of
Œfixedness‹. We approach sentences in English with the contextual
assumption that the first verb will be the main verb, barring<j <
contrary semantic or syntactic information (viz., Milne, 1982).
If "sailed" is assumed to be the main verb, then we do not know
what to do with the verb "sank". But "sailed" may also be the
verb of a Œsubordinate clause‹, as in the following examples:
(a) The ship Œsailed‹ by the commodore Œwas‹ a beautiful sight.
(b) The ships Œsailed‹ at Newport Œare‹ racing sloops.
(c) To my surprise, a ship Œsailed‹ by a good crew Œsank‹.
Here the main verbs always come later in the sentence. The
trouble with sentence (1) is that we tend to become committed to
one syntactic interpretation before all the evidence is in, and
we may find it impossible to back away from it. In the most
general terms, we are captured by one unconscious interpretation
of the beginning of the sentence --- we are Œfixated by the wrong
syntactic context‹.
Fixedness can be found in all kinds of problem solving. It
is found in vision, language perception, in solving puzzles, in
science, literature, politics, and warfare (Luchins, Duncker,
Levine; Bruner & Potter). American policy during the Vietnam War
may have been an example of fixedness, since it followed certain
assumptions about international relations that were widely
accepted at that time, across the political spectrum. In
retrospect, some of those assumptions are questionable. But that
is just the point about fixedness: seen in retrospect or from
"the outside", it is hard to believe that the fixated person
cannot see the "obvious" solution. But Œwithin‹ the fixating
context the solution is not obvious at all: it is literally
impossible to perceive.
Yet fixedness is a completely normal part of learning.
Whenever we try to learn something before we have the knowledge
needed to make sense of the material, we may find ourselves
interpreting it in the wrong context. McNeill (1966) cites the
example of a mother trying to teach her child something about
English negation --- a bit prematurely:
ŒChild‹: Nobody don't like me.
ŒMother‹: No, say "Nobody likes me."
ŒChild‹: Nobody don't like me.
ŒMother‹: No, say "Nobody likes me."
Œ(Eight repetitions of this dialogue.)‹
ŒMother‹: No, now listen carefully, say, "Nobody likes me".
ŒChild‹: Oh! Nobody don't likes me.
A year later the same child would laugh at the error, but<j <
when the dialogue was recorded he or she was not prepared to
perceive the difference. In learning, as in life, readiness is
all.
A major point is to realize that our notion of "fixedness"
depends critically on having an outside point of view in which
the mistake Œis‹ a mistake. That is to say, as adults we can find
the above example comfortably amusing, because we know the right
answer. But for the child the error is no error at all. The
"flawed" sentence is not experienced as erroneous; in terms of
the child's internalized rules, it is not an error at all.
ŒSelective attention as a contextual fixedness effect.‹
One powerful implication is that "fixedness" exists in
states of mind that we consider to be perfectly correct. For
example, one can plausibly argue that selective attention is a
fixed state of mind --- after all, in shadowing speech in one ear
we are utterly oblivious to the unattended stream of speech, as
much as the child in the language example is oblivious to the
"correct" sentence. Thus the remarkable ability of one stream of
speech to capture our conscious experience to the exclusion of
any other looks like a contextual fixedness effect. Notice that
structural similarities between the two streams of speech will
cause leakage between them; that is, when they share context, the
"unconscious" stream tends to affect the conscious stream (e.g.
Norman, 1976). Normally we can hear the acoustical qualities of
the unattended ear, perhaps because these qualities match the
acoustical contexts of the attended ear. After all, the attended
ear must detect a range of sounds as well. Further, when the
semantic context of the attended involves an ambiguous word like
"bank," it is open to influences from the unattended ear to the
extent those influences are consistent with the semantic
ambiguity (MacKay, 1973). In Table 4.1 this point is made by
listing "acontextual" information on the unconscious side of the
contrastive table. When there is potentially conscious input, but
the right context is not brought to bear on it, it does not
become conscious.
Similarly, in absorbed states of mind --- in reading a fine
novel or watching an entrancing motion picture --- we are deaf
and blind to the world. In absent-minded states we are likewise
captured by one train of thought to the exclusion of others
(Reason, 1983). One plausible supposition is that all these
states are initiated, shaped, and bounded by powerful context
hierarchies that permit no interruption for the time being.
Only a change in the fixating context, or giving up on the
task, can release us from fixedness. Above, this change in
context is created by the little story about Napoleon's ships
running the English blockade. This creates a new context that<j <
works, but that no doubt has its own fixating properties. This is
the normal case, of course: we may change to more a effective
context, but we cannot undo context as such. Inevitably we are
condemned to both its advantages and drawbacks.
The existence of fixedness provides extensive evidence for
the power of contexts. Next, we consider the case in which
context actually enters into conscious experience.
4.13 Top-down influences and the pervasiveness of
ambiguity.
Many domains of experience are full of local ambiguities.
This is obvious in some cases, and not so obvious in others. For
some obvious examples, there are many times when information
about the world is degraded, inadequate, or forgotten. Examples
include listening to a conversation in a noisy room, trying to
see an oncoming bus at a great distance, or walking through a
dark room at night. In all these cases we rely more than usual on
the inner context to constrain conscious experience. In the
social realm, it is terribly important for us to know other
people's minds --- their intentions, beliefs, and attitudes
toward us. But we cannot read their minds directly. The evidence
we have is ambiguous, and hence vulnerable to our own goals and
expectations, wishes and fears. We often make inferences about
other people's minds with a degree of confidence that is simply
not justified by the evidence (Nisbett & Wilson, 1977). In this
case, inner context controls our experience far too often.
Political convictions show this even more graphically. A glance
at the editorial pages of a newspaper shows how people with
different convictions use the same events to support opposite
beliefs about the world, about other people, and about morality.
Or take the domain of "the future". Human beings are intensely
concerned about the future, and we often have strong beliefs
about it, even when future events are inherently probabilistic
(Tversky & Kahneman, 1973). The evidence is inadequate or
ambiguous, and hence we rely more and more on internal contextual
constraints.
Those examples are fairly obvious, but there are many
ambiguous domains in which we experience events with great
confidence, though careful experiments show that there is much
more local uncertainty than we realize. There is extensive
evidence that our own bodily feelings, which we may use to infer
our emotions, are often ambiguous (Schachter & Singer, 1962;
Valins, 19xx). Further, our own intentions and reasons for making
decisions are often inaccessible to introspection, or at least
ambiguous (Nisbett & Wilson, 1977) (see Chapters 6 and 9). Our
memory of the past is often as poor as our ability to anticipate
the future, and it is prone to be filtered through our present
perspective (Bransford, 1979; Mandler, 1985). Historians must<j <
routinely cope with the universal tendency of people to reshape
the past in light of the present, and lawyers actively employ
techniques designed to make witnesses change their memory of a
crime or accident (see Bransford, 1979).
Even perceptual domains that seem stable and reliable are
actually ambiguous when we isolate small pieces of information.
Every corner in a normal rectangular room can be interpreted in
two ways, as an outside or an inside corner. To see this, the
reader can simply roll a piece of paper into a tube, and look
through it to any right-angled corner of the room. Every room
contains both two and three-dimensional ambiguities in its
corners, much like the Necker cube and book-end illusions (Figure
2.1). Similarly, the experienced brightness of surfaces depends
upon the brightness of surrounding surfaces (Gelb, 1932;
Gilchrist, 1977). Depth perception is controlled by our
contextual assumptions about the direction of the incoming light,
about the shape and size of objects, and the like (Rock, 1983).
These ambiguities emerge when we isolate stimuli --- but it is
important to note that in normal visual perception, stimulus
input is often isolated. In any single eye fixation we only take
in a very small, isolated patch of information. Normal detailed
(foveal) vision spans only 2 degrees of arc; yet when people are
asked about the size of their own detailed visual field, they
often believe it must be about 180 degrees. Even the visual
world, which seems so stable and reliable, is full of local
ambiguities (Marr, 1982).
Language provides a great many examples of ambiguity.
Indeed, every level of linguistic analysis has its own kind of
ambiguity. Thus,
1. Ambiguities of sound. The English /l/ is perceived as
either /r/ or /l/ by Japanese speakers, while the unaspirated /k/
(as in "cool") is freely exchanged by English speakers with the
aspirated /kh/ (as in "keel"). In Arabic this difference marks
very different words. Most English speakers simply do not hear
the tones that are critical in languages like Chinese. Further,
there are many identical strings of sounds in every language that
are divided up differently, as in "ice cream" and "I scream" in
English. We typically become conscious of these ambiguous sound
sequences only in learning a new language.
2. Morphemic ambiguity. The final /s/ in English has four
different morphemic interpretations. It can be plural (the
books), third person singular verb (he books the tickets),
possessive (the book's cover), and plural possessive (the books'
covers).
3. Lexical ambiguity. A glance at the dictionary should
convince anyone that each word has more than one meaning. More
common words tend to have more meanings.
4. Syntactic ambiguity. There are numerous syntactic<j <
ambiguities. The best-known ones are the surface and deep-
structure ambiguities of Chomskyan theory (Chomsky, 1957, 1965).
Thus, "old men and women" is a surface ambiguity that involves
grouping: one can have "old (men and women)", or "(old men) and
women". Sentences like "Flying planes can be dangerous" and "They
are eating apples" have ambiguities that cannot be represented in
a single tree diagram; they involve ambiguity in underlying
subjects and objects.
5. Discourse ambiguity. Consider the following example:
(a). The glass fell off the table.
(b). It broke.
(b'). It was always a little unstable.
The referent of "it" changes between (b) and (b'). It can
only be determined by an appeal to context, and to the subject's
knowledge about glasses and tables. Such ambiguities are
extremely common.
6. Referential ambiguities. This occurs when we refer to
"that chair" in an auditorium full of chairs, or to "that book"
in a library.
7. Semantic ambiguity. All too often, concepts do not relate
clearly to other concepts. What really is consciousness? What is
an atom, or a physical force, or a biological species? All
unresolved scientific questions involve deep semantic ambiguity.
8. Topical uncertainty and ambiguity. Consider the following
paragraph (Bransford, 1979):
The procedure is actually quite simple. First you arrange
items into different groups. Of course one pile may be sufficient
depending upon how much there is to do. If you have to go
somewhere else due to lack of facilities that is the next step;
otherwise, you are pretty well set. It is important not to overdo
things. That is, it is better to do too few things at once than
too many. In the short run this may not seem important but
complications can easily arise. A mistake can be made as well. At
first, the whole procedure will seem complicated. Soon, however,
it will become just another facet of life. It is difficult to
foresee any end to the necessity for this task in the immediate
future, but then, one never can tell. After the procedure is
completed one arranges the materials into different groups again.
Then they can be put into their appropriate places. Eventually
they will be used once more and the whole cycle will have to be
repeated. However, that is part of life.
Confused? Here is the context: the paragraph is about
"washing clothes". If you read it again, it will be much more
comprehensible, details will clarify in experience, and memory
for the material will improve greatly. <j <†
What is the point of this litany of ambiguities? It is that
ambiguity is pervasive; but the Œconscious experience of ambiguity‹
is quite rare. We generally gain information about a world that
is locally ambiguous, yet we usually experience a stable,
coherent world. This suggests that Œbefore input becomes
conscious, it interacts with numerous unconscious contextual
influences to produce a single, coherent, conscious experience.‹
Consciousness and context are twin issues, inseparable in the
nature of things.
As we pointed out in Chapter 2, the Global Workspace
architecture was originally developed to deal precisely with the
problem of unifying many ambiguous or partial sources of
information into a single, unified solution (2.xx).
Another source of evidence on contexts is considered next.
4.14 Decontextualization: Strong violations of context can
become consciously accessible.
Unconscious contextual assumptions can become consciously
accessible. Every statement we hear or read has presupposed
(contextual) information that must be understood before it can
make sense. The "washing machine" paragraph shown above is an
example. But these contextual presuppositions remain unconscious
Œunless‹ they are violated. If we suddenly speak of putting a five-
pound weight into the washing machine, we tend to become
conscious of one contextual assumption. We will call this process
decontextualization. It is a theoretically central phenomenon.
Consider the following little story, which is quite normal, so
that presupposed ideas tend to remain unconscious:
(1) "It was a hot day. Johnny walked to the store and bought
some ice cream to eat. Then he brought home a snow cone for his
mother."
But now consider the following version. (It will help to
read slowly):
(2) a. It was a hot day in December.
b. Johnny walked three hours to the town store.
c. He was completely broke but not very hungry.
d. He bought a gallon of ice cream and ate to his
heart's content.
e. He also brought a snow cone home to his mother.
--------------------------------------------
FIG. 4.14 THE PRESUPPOSITION OF "BUYING" <j
† THAT MAY BECOME CONSCIOUS UPON VIOLATION.
--------------------------------------------
The second story makes implausible a number of conditions in
the presupposed context of the event "buying ice cream", so that
we find this normally unconscious knowledge becoming conscious.
Story (1) probably left unconscious the fact that "walking to the
store to get ice cream" is usually assumed to take a matter of
minutes, not three hours; that hot days are likely in summer, not
in December; that buying involves money, and thus "being broke"
excludes buying anything; that one can eat ice cream even without
being very hungry, but that one would certainly not eat a gallon
of ice cream if one were not very hungry; and that walking home
for three hours on a hot day, carrying a snow cone, would cause
it to melt. When these implausible claims are presented, the
contradicted context tends to come to mind spontaneously.
We can think of this presupposed context as a set of
Œstable, predictable constraints‹ on normal discourse (Foss, 1982;
Clark & Carlson, 1981). As long as a piece of presupposed
knowledge remains predictable it also tends to remain
unconscious. But when it is strongly violated, its consequences
tend to become conscious in some way. This is similar to the
pattern we find with violated expectations in stimulus
habituation, with obstacles created in automatic tasks, and with
increases in task difficulty for habituated mental images (see
1.x; 5.xx). In all these cases novel information is created by
the violation of established, predictable properties of the
situation. Of course, when we make parts of a conceptual context
conscious, this places the newly conscious material in its Œown‹
unconscious context. Becoming conscious of contextual knowledge
while it is acting as context is like chasing one's tail; it is
ultimately impossible.
4.15 A summary of the evidence for unconscious contexts.
First, conscious experiences change later, related
experiences long after the earlier ones have become unconscious.
Presumably, the first experience creates a context within which
the later one is shaped and defined.
Second, the universal phenomenon of fixedness suggests that
all conscious and deliberate processes are bounded by assumptions
that are unconscious to the subject, though they may be obvious
to outside observers. Selective attention and aborbed states may
be variations on this theme of fixedness.
Third, there is extensive evidence for local ambiguity in
all areas of life --- the past, the future, other people's minds,<j <
our own feelings, visual and auditory perception, language
understanding, etc. All these domains are rife with ambiguity.
Yet ambiguities are rarely experienced Œas‹ ambiguities. Normally,
many different contexts interact to create a single conscious
interpretation of reality.
Fourth, strong violations of our contextual assumptions can
become conceptually conscious --- that is, we can refer to these
surprises as objects of experience, and by becoming conscious, we
can sometimes change our previous contextual way of thinking
about them.
We can assess contexts in two convenient ways. First,
priming tasks can be designed to be sensitive to the information
contained in a dominant context. Current cognitive psychology has
dozens of examples of the use of such priming techniques (e.g.
Swinney, 1979; Baars, 1985). Second, one can observe the
occurrence of surprise to events that violate contextual
expectations. Thus changes in heart rate --- a measure of
surprise --- have been used to assess the existence of phoneme
boundaries in infants who could not possibly tell us about their
experience. Surprise could be used much more often with adults,
because there is little reason to think that adult voluntary
report of contextual structures is accurate; hence we may miss
contextual events in adults rather often, because of our reliance
on verbal report.
4.2 Several kinds of contexts.
We can talk about several kinds of contexts: First, the
context of perception/imagery; next, the context of conceptual
thought; third, goal-contexts, which evoke and shape actions; and
finally, the context of communication that is shared by two
people talking with each other, or by ourselves talking to
ourselves. Notice that some of these contexts actually shape
conscious experience as such, while others evoke conscious
thoughts and images, or help select conscious percepts.
Perceptual/imaginal contexts clearly enter into the conscious
qualitative experience. A goal context may simply serve to recall
a word (6.0) or evoke a mental image. That is, not all contexts
necessarily enter into the experience itself. Naturally these
different kinds of contexts interact with each other. Perceptual
events and images have a lot of influence on conceptual thinking;
concepts influence inner speech, images, and the selection of
perceptual events; goals influence concepts, and Œvice versa‹. We
now examine the types of context in a little more detail.
<j <† 4.21 The Context of Perception and Imagery.
ŒThe Context of Perception.‹
Imagine sitting in a tiny, well-lit movie theater, looking
at a metallic disk instead of a movie screen. The disk appears
to be white. But now someone lights a cigarette, and as the smoke
curls upward you see it floating through a slender but powerful
light beam, coming from the rear of the theater, and aimed
precisely at the metal disk. You look back at the disk, and
suddenly notice that it isn't white at all, but black. This is
the Gelb Effect (Gelb, 1932), and can be summarized by saying
that the color of a surface is a function of the perceived
incoming light. If we are never conscious of the incoming light,
we will attribute the brightness of the disk to its surface color
and not to the light. Once having seen the cigarette smoke
intersecting the light beam, the disk is seen to be black.
Similarly, if we turn a picture of the moon's craters upside
down, the experience of depth is reversed, so that craters are
seen as hills. This is because scenes are interpreted under the
assumption that light comes from above, as indeed it usually
does. When the photo of the moon is turned upside-down, the light
is still assumed to come from the top of the picture, and
concavities are turned into convexities (Rock, 1983).
Perceptual research since the 19th century has uncovered
hundreds of such phenomena. All of them can be summarized by
saying that complex and subtle unconscious systems, which we call
contexts, shape and define conscious perceptual experiences.
ŒThe Context of Imagery.‹
Imagery has not been studied as extensively as perception,
but over the past decade very interesting findings have emerged,
suggesting constraints on visual imagery of which we are
generally unconscious. These constraints tell us both about the
Œformat‹ and the Œcontent‹ of imagery (Kosslyn & Schwartz, 1981).
The "field" of visual imagery has a close resemblance to vision:
it has the same flat elliptical shape as the visual field, it
presents us with one perspective on a potentially three-
dimensional spatial domain, and the scanning time needed to move
From one point to another in the Mind's Eye is a linear functi
of the distance between the two points, just as we might expect
of the visual field.
Clearly as we learn more about mental imagery, we will
continue to find more of these constraints, which are largely
unconscious until they are brought to mind. <j <†
4.22 The Conceptual Context: conceptual presuppositions
are normally unconscious.
Anyone who has tried to think very clearly about some topic
must know from experience that our Œstable presuppositions‹ tend to
become unconscious. Whatever we believe with absolute certainty,
we tend to take for granted. More accurately perhaps, we lose
sight of the fact that Œalternatives to our stable presuppositions
can be entertained‹. Indeed, scientific paradigm shifts generally
take place when one group of scientists begins to challenge a
presupposition which is held to be immutable (and hence is
largely unconscious) in the thinking of an older scientific
establishment. In his autobiography, Albert Einstein described
this phenomenon in 19th century physics (1949):
"... all physicists of the last century saw in classical
mechanics a firm and final foundation for all physics, yes,
indeed, for all natural science ... Even Maxwell and H. Hertz,
who in retrospect appear as those who demolished the faith in
mechanics as the final basis of all physical thinking, Œin their
conscious thinking‹ adhered throughout to mechanics as the secured
basis of physics" (p. 21; italics added).
Some pages later he recalls how he gained the insight which
led to the Special Theory of Relativity:
"... After ten years of reflection such a principle resulted
From a paradox upon which I had already hit at the age
sixteen: If I pursue a beam of light with the velocity Œc‹ (the
velocity of light in a vacuum), I should observe such a beam of
light as a spatially oscillatory electromagnetic field at rest.
However, there seems to be no such thing ... One sees that in
this paradox the germ of the special relativity theory is already
contained. Today everyone knows, of course, that all attempts to
clarify this paradox satisfactorily were condemned to failure as
long as the axiom of the absolute character of time, viz., of
simultaneity, Œunrecognizedly was anchored in the unconscious‹. (p.
53; italics added.)
Kuhn (1962) quotes Charles Darwin to much the same effect:
"Darwin, in a particularly perceptive passage at the end of
his ŒOrigin of Species‹, wrote, 'Although I am fully convinced of
the truth of the views given in this volume ... I by no means
expect to convince experienced naturalists whose minds are
stocked with a multitude of facts all viewed, during a long
course of years, from a point of view directly opposite to mine.
... (B)ut I look with confidence to the future --- to young and<j <
rising naturalists, Œwho will be able to view both sides of the
question‹ with impartiality.'" (Italics added.)
Darwin observed that many older naturalists were simply
unable to consciously think consistently of the alternatives to
their own, stable presuppositions. In psychology this phenomenon
is easy to observe today, since the field has recently passed
through something much like a paradigm shift (see Baars, 1986a).
It is still remarkably easy to find psychologists who find it
impossible to take the existence of consciousness seriously. For
these people, the implications of consciousness as something
scientifically real and important remain hidden and unconscious.
It would be interesting to find out why conceptual
presuppositions tend to become conscious so readily in a simple
story like the one cited above (), compared to the case of
scientific change, where Darwin, Einstein, and many others have
complained so much about the inability of other scientists to
entertain alternatives to their own presuppositions. Was it
because these scientists were emotionally invested in their
customary way of viewing the world? Or do more complex knowledge
domains make it more difficult to see contextual alternatives and
their consequences? Or both?
ŒPerceptual vs. conceptual contexts‹
There are some interesting differences between the
perceptual and the conceptual context. In the case of perception,
when the context is challenged we do not "perceive" the challenge
directly, though we can conceptualize it. That is, in getting
used to the swaying of a small sailing boat, we can think of the
conceptual fact that the horizon is really not swaying, but the
boat is. It is not clear whether this conceptual realization
about the perceptual context helps the process of adaptation.
The Ames room provides another good example. As we noted
above, a room with trapezoidal walls can appear normal so long as
the observer is stationary and monocular. In that case, a person
walking in the room will be seen to grow as he or she is
approaching the low end of the trapezoidal wall, and to shrink in
the opposite direction. The visual system, forced to choose
between revising its assumptions about the room or about the
constant height of people, prefers to let human height change to
keep the room the same. However, when the observer is allowed to
bounce a few ping-pong balls against surfaces in the room, the
challenge to contextual assumptions becomes overwhelming. The
balls bounce off at odd angles for a normal room; they take
longer to reach the opposite wall when it is farther away, even
though it seems close by. Now the observer experiences a shift:
the room is seen to be truly trapezoidal, and human height is
experienced as constant. Previous unconscious assumptions are<j <
revised, and we now experience the room veridically.
Things are quite different in solving a conceptual problem.
Einsteinian physicists who began to question the time axiom of
Newtonian theory were able to change their presuppositions
directly, voluntarily, even though their more traditional
colleagues found this difficult or impossible. In conceptual
contexts, we can at times make a piece of context consciously
accessible, and change it. The new conceptual context then begins
to shape the interpretation of scientific observations. Notice
that when we question a presupposed idea, it is no longer
presupposed, but focal and conscious. It is therefore interpreted
in its own conceptual context. When we Œtalk‹ about our conceptual
context we can make a piece of it conscious.
ŒScientific paradigms as largely unconscious contexts.‹
Communication problems occur when people try to exchange
ideas under different contextual assumptions. This is especially
clear in the case of Œparadigmatic differences‹ in a scientific
community. One might expect science at least to be free of such
communication problems, because scientists deal with a shared,
observable empirical domain, and because mature sciences make use
of explicit formal theories. Not so. Historians have long
remarked on the frequency of communication problems in science,
but it is only with Thomas Kuhn's seminal monograph ŒThe structure
of scientific revolutions‹ (1962) that these communication
problems have come to be widely acknowledged as part of the
fundamental nature of science. Kuhn described two kinds of
evolution in the history of science: within a certain framework,
or "paradigm", development is cumulative, since scientists share
common tools, goals, typical problems, and assumptions about
reality. Thus physics enjoyed a shared paradigm in the two
centuries after Newton's ŒPrincipia Mathematica‹ until the late
19th century, when the paradigm began to develop difficult
internal contradictions. Einstein's relativity theory solved some
of those problems, giving rise to a new framework within which
physicists could again communicate without serious problems for
some time; but Einsteinian physicists had great difficulty
communicating with those who continued to view the world in
Newtonian terms. Kuhn calls this phenomenon "the
Œincommensurability‹ of competing paradigms":
"Since new paradigms are born from old ones, they ordinarily
incorporate much of the vocabulary and apparatus, both conceptual
and manipulative, that the traditional paradigm had previously
employed. But they seldom employ these borrowed elements in quite
the traditional way. Within the new paradigm, old terms,
concepts, and experiments fall into new relationships with the
other. The inevitable result is what we must call, though the
term is not quite right, a Œmisunderstanding‹ between the two<j <
competing schools. The laymen who scoffed at Einstein's general
theory of relativity because space could not be "curved" --- it
was not that sort of thing --- were not simply wrong or mistaken.
Nor were the mathematicians, physicists, and philosophers who
tried to develop a Euclidian version of Einstein's theory. What
had previously been meant by space was necessarily flat,
homogeneous, isotropic, and unaffected by the presence of matter.
If it had not been, Newtonian physics would not have worked. To
make the transition to Einstein's universe, the whole conceptual
web whose strands are space, time, matter, force, and so on, had
to be shifted and laid down again on nature whole...
ŒCommunication across the revolutionary divide is inevitably
partial.‹ Consider, for another example, the men who called
Copernicus mad because he proclaimed that the earth moved. They
were not either just wrong or quite wrong. Part of what they
meant by 'earth' was fixed position. ŒTheir‹ earth, at least, could
not be moved. Correspondingly, Copernicus' innovation was not
simply to move the earth. Rather, it was a whole new way of
regarding the problems of physics and astronomy, one that
necessarily changed the meaning of both 'earth' and 'motion'.
Without those changes the concept of a moving earth Œwas‹ mad. ...
"These examples point to the ... most fundamental aspect of
the Œincommensurability between competing paradigms.‹ In a sense
that I am unable to explicate further, the proponents of
competing paradigms practice their trades in different worlds.
One contains constrained bodies that fall slowly, the other
pendulums that repeat their motions again and again. In one,
(chemical) solutions are compounds, in the other mixtures. One is
embedded in a flat, the other in a curved, matrix of space.
Practicing in different worlds, the two groups of scientists see
different things when they look from the same point in the same
direction. Again, that is not to say that they can see anything
they please. Both are looking at the world, and what they look at
has not changed. But in some areas they see different things, and
they see them in different relations one to the other. That is
why a law that cannot even be demonstrated to one group of
scientists may occasionally seem intuitively obvious to another.
Equally, it is why, Œbefore they can hope to communicate fully,
one group or the other must experience the conversion we have
been calling a paradigm shift‹. Just because it is a transition
between incommensurables, the transition between competing
paradigms cannot be made a step at a time, forced by logic and
neutral experience. ... (Kuhn, 1962/1970, pp. 149-151;
italics added.)
Why is it so difficult for committed scientists to change
paradigms? From our model, it would seem that change is hard, at
least in part because at any single moment the bulk of a paradigm
is unconscious. In our terms, paradigms are conceptual contexts.
If one tried to make a paradigm conscious, one could only make
one aspect of it conscious at any one time because of the limited
capacity of consciousness. But typically paradigm-differences
between two groups of scientists involves not just one, but Œmany‹<j <
different aspects of the mental framework simultaneously. This
may also be why conversion phenomena in science (as elsewhere)
tend to be relatively rapid, all-or-none events that seem to have
a not-quite-rational component. In fact, Kuhn compares the
experience of conversion to a "Gestalt switch" such as we observe
with the Necker Cube (2.1, 4.11).
4.23 Intentions as Goal Contexts.
Thus far we have talked about two kinds of context, the
qualitative (perceptual-imaginal) context and the conceptual
context. Conscious experiences also interact with a third kind of
unconscious context, which we will call Œthe Goal Context‹. Goal
Contexts are useful in understanding problem solving, intentions,
and voluntary control. We will postpone a detailed consideration
of Goal Contexts until Chapters 6 and 7. However, it is important
at this point to introduce the concept of an ordered Œgoal
hierarchy‹ --- simply, the idea that goals are ordered in
significance at any point in time, and that higher (more
significant) goals will tend to predominate over lower ones. This
is by no means a new idea; it has been suggested by numerous
motivational and cognitive psychologists (e.g. Maslow, 1970), and
the computational implications of goal hierarchies have been
worked out in some detail by artificial intelligence researchers.
Our emphasis at this point is on the control of conscious events
by such contextual goal hierarchies, which are diagrammed in
Figure 4.23.
------------------------------
INSERT FIGURE 4.23 about here.
-----------------------------
Note that goal hierarchies cannot be rigid over time. For
instance, the goal of eating must rise higher in the hierarchy
after food deprivation. But at any one moment the hierarchy
is ordered. That is, at any particular time we will prefer food
over sex over watching TV. Some goals are more stable over time:
generally survival has a higher priority than avoiding boredom.
We will not develop this notion here; we simply suggest that
there a set of ordered goals act as context for the flow of
conscious experience. The evidence for this claim is developed in
Chapters 6 and 7.
4.24 Other types of context.
Social and cultural contexts usually operate unconsciously.
The sociologist Ervin Goffman writes, "When the individual in our
Western society recognizes a particular (social) event, he tends
...(to) employ one or more frameworks or schemata of
interpretation ... to locate, perceive, identify, and label a
seemingly infinite number of concrete occurrences defined in its
terms. ŒHe is likely to be unaware of such organized features as
the framework has and unable to describe the framework with any
completeness if asked,‹ yet these handicaps are no bar to his
easily and fully applying it." (Goffman, 1974, p. 21; Italics
added.) Anthropologists often encounter their own cultural
presuppositions in a dramatic way when they enter a culture that
violates those presuppositions; as usual, unconscious
presuppositions can become conscious when they are severely
violated. A member of another culture may seem to thrust his face
toward a Westerner in a conversation at an unacceptable eight
inches away. This experience may be shocking or offensive, but it
makes conscious what is normally taken for granted: namely the
fact that we, too, adopt a typical social distance. Thus
unconscious customs and habits come to the foreground. Custom
leads to adaptation and loss of consciousness; this is why
children, novices, and strangers can guide us to become conscious
again of things we have lost touch with in the process of
becoming adults, experts, and members of various in-groups.
These properties of context have major implications for sociology
and anthropology. For instance, all cultures have periodic
ceremonies, festivals, and initiation rites using dramatic or
even traumatic symbolism; a major function of these events may be
to create and renew memorable conscious experiences that invoke
and reinforce the unconscious contextual assumptions of the
society.
Much the same sort of thing is true for motivation and
personality research. The genius of Freud has led us to believe
that much of our lack of awareness of ourselves is due to
repression. This may be true, but even without repression we
should be ignorant of regularities in our own actions, beliefs
and experiences, simply because regularity by itself creates
unconscious context (e.g. Nisbett & Wilson, 1977; see Chapter 9).
It is possible that motivational mechanisms like repression and
denial make use of contexts to ward off painful or confusing
experiences (9.xx).
Many kinds of knowledge can be profitably viewed in these
terms. One of the most important is the Œcontext of communication‹.
Indeed, we need some sort of shared context to conduct any social
relations. To communicate we must share with each other
unconscious assumptions about ourselves, our audience, our
conceptual framework, and the perceptual world. The communicative
context is a close relative of the Conceptual Context, but it is<j <
not identical. Here we can follow Clark and Carlson (1981), who
define "the intrinsic context for understanding what a speaker
means" as "the common ground that the Œlistener‹ believes holds at
that moment between the speaker and the listeners" (italics
added). Clearly the speaker's and the listener's context need
not be identical, though they should overlap considerably if
communication is to succeed. We would emphasize that this shared
set of beliefs is probably unconscious at the time the
participants hold it, although they will consciously recognize
violations of the implicit context. Clark and Carlson cite formal
demonstrations that common ground is necessary in social
conventions, in speech acts, and in definite reference. Much the
same can be argued for the "given-new contract," the agreement
between two speakers to focus on new material, and to take shared
givens for granted (e.g. Clark & Clark, 1977; Chafe, 1970).
When two people know each other very well they can often
communicate with remarkable brevity. A word, a glance, an
unexpected silence --- these can often say volumes. The reason
for this economy of expression is clear: people who talk together
for many years share so much context that very little needs to be
made explicit.
An elegant experiment by David R. Olson (1970) serves as a
prototype for the communicative context. Olson asked one child to
tell another child, hidden behind a screen, where to find a gold
star located beneath a white, round block. What the first child
said depended not just on the block referred to, but also on the
other blocks which were present. If the other blocks were also
white but had different shapes, the child would refer to the
Œdisambiguating shape‹ by saying, "It's under the round one". If,
on the contrary, the colors were different but the shapes were
all the same, the child would say, "It's under the white one" ---
referring to the Œdisambiguating color‹. Thus the description of
the object differed according to the Œcontext of alternatives
which were assumed to exist in the mind of the listener.‹ This is
apparently a general property of communication, one that applies
to adults and to very different semantic domains as well.
Most listeners and speakers already share a tremendous
amount of context: the children in Olson's experiment share a
great deal of knowledge about the world of objects, about the
size, shape, color, weight, and playability of blocks, about gold
stars, about language, and even about each other's knowledge
about these things. Most of this context can be taken for
granted; it does not need to be mentioned, nor does it need to be
made conscious for communication to work. Only the information
needed to disambiguate Œrelevant choices‹ in the shared context
needs to be specified. This is why people who share a great deal
of context seldom need to make it explicit.
<j <† ŒContexts in communicating with ourselves.‹
The more context we share with other people, the less we must
make conscious and explicit. This observation suggests something
about our Œinner‹ dialogue as well. Having lived with ourselves for
so long, it seems likely that we can communicate to ourselves
with minimal conscious inner speech; and each conscious thought
can have reference to voluminous amounts of knowledge. We can
hypothesize that what needs to be made explicit in communication
is closely analogous to what needs to be made conscious in the
mind of the speaker and listener (Chafe, 1970). That is, in
general it may be true that we need to become conscious only of
information which disambiguates some relevant context of
alternatives, even in our own minds --- everything else can
remain unconscious. This view is closely related to the notion
that conscious events are Œinformative‹, i.e. that they select one
interpretation from a larger context of alternative
interpretations (see Chapter 5).
4.25 The different kinds of context interact.
Perception and imagery are key ingredients in the Conceptual
and Goal Context. In particular, conceptual thinking is affected
by inner speech and by visual images --- both of which are
controlled by the Context of Imagery. Indeed, as Rosch and her
associates have shown, much of our abstract thinking is heavily
swayed by imageable "prototypes" (). The class of birds is
represented not so much by an abstract description of birds as a
biological genus, but rather by some particular imageable bird,
like a robin, that stands for the abstract class. Similarly, the
class of chairs is often mentally represented by the classical
kitchen chair, made of wood, with a square back and seat, and
with the natural wood grain showing through the lacquer. This
prototypical chair is neither the average chair we encounter, nor
is it an adequate abstract description of all chairs. Rather, it
is something we can conveniently imagine consciously.
Prototypical images serve to index abstract descriptions that
cannot be visualized.
Similarly, we know that abstract thinking is heavily
influenced by metaphors, which can usually be imaged, but which
stand for more abstract things. Lakoff and Johnson (1980) discuss
a number of such common metaphors, such as "memory is like a
container," "love it like a journey," "electricity is like a
stream," or "the atom is like a solar system." These metaphors
influence thinking, sometimes far more than they should --- a
useful caution in this book!
Thus the Conceptual Context is heavily influenced by the<j <
Contexts of Perception and Imagery. But influence runs the other
way as well. Perception is relatively impervious to conceptual
thought --- try as we might, we cannot change the visual world by
thinking about it --- but conceptual thinking can evoke different
visual images and inner speech. Conceptual processes may also
select perceptual events to pay attention to. In later chapters
we will argue that imagery and perception have great influence on
action goals as well (6 and 7).
In sum, different kinds of context interact. Especially
significant is the way in which conceptual thinking and goals are
triggered by, and in turn influence, the Context of Perception
and Imagery. In Chapters 6 and 7 this point will suggest an
answer to the question we asked in Chapter 1: What is the
relationship between qualitative conscious contents, like
percepts and mental images, and non-qualitative "conscious"
beliefs, concepts, expectations and intentions?
4.3 Modeling contextual knowledge.
We can apply the familiar facts about cognitive
representations to contexts. Piaget's description of schemata,
cognitive views about scripts, semantic networks, organization in
memory, story grammars, currently activated knowledge and the
like, all these statements can be applied to contexts, as long as
the knowledge structure in question is generally unconscious
while influencing conscious contents (Piaget, Norman & Rumelhart,
Bransford, etc.). This distinctive fact makes a great difference.
We will use a notational convention (Figure 4.3x) in which
contexts are shown as horizontal "frames", with horizontal length
symbolizing duration in time, and vertical length representing
cooperation or competition with respect to other contexts and
with the Global Workspace. In general, "higher" contextual frames
imply a more encompassing context that is presupposed by lower
ones. Contextual frames are nested, and the higher ones "embrace"
the lower contexts. Of course our diagrams are schematic only.
Details of contexts must be worked out in every domain. But here
we are concerned mainly with the general properties of contexts.
-------------------------------------------------
INSERT FIGURE 4.3 ABOUT HERE
-------------------------------------------------
4.31 Contexts as stable coalitions of specialized processors. <j <†
So far our models have had only two entities: specialized
processors, which are believed to be unconscious, and a global
workspace, whose contents are conscious. We now add a third
construct, a "context", a knowledge structure that is
unconscious, but which constrains whatever becomes conscious.
How do contexts relate to the specialized processors
discussed in previous chapters? After all, specialized processors
are unconscious also. In fact, we can treat a context as a
cooperating group of specialized processors with ready access to
the Global Workspace. If contexts are to shape conscious
experience they obviously must be able to interact with GW
messages quickly and easily. Some of this interaction may
conceivably take place even before global messages reach the
global workspace. But the arguments given in Chapter 2 about the
ability for any ambiguity to be resolved by any other knowledge
source, given enough time and learning, suggest that some
contextual systems must interact with potentially conscious
events through the global workspace (2.xx). We will refer to the
set of currently operative contexts as the Current Dominant
Context Hierarchy, or Dominant Context for short (viz., Shallice,
1978). Any group of specialized processors that can cooperate to
shape conscious experience, and that has a routine, committed way
of processing information, will tend to behave like a context.
Figure 4.31 makes this point simply by showing that unconscious
processors together can look like a contextal frame. Another way
of saying this is that contexts involve unconscious processors
that are Œalready committed‹ to a certain stable way of processing,
and that tend to shape global messages to their own committed
organization. Contexts presumably develop over time, in a process
coalition-formation and competition, in roughly the same way that
a majority party in a legislature evolves a working, stable
coalition between among its members.
---------------------------------------------
Insert Figure 4.31 about here.
CONTEXTS AS STABLE COALITIONS OF SPECIALISTS.
PUT CONtext Hierarchy in the same fig?
---------------------------------------------
4.32 The current Dominant Context imposes unconscious
constraints on what can become conscious.
<j <† Just by reading and understanding the foregoing sections,
the reader will have gained a partial framework, or, well ... a
Œcontext‹ for the rest of this chapter. This suggests that you will
experience the following material in a different way, Œeven when‹
you are not consciously recalling the earlier material. In this
way do we humans continuously define and redefine reality.
Contexts are not conscious at the time when they have their
influence, though they might be conscious a second, an hour, or
half a lifetime before the experience which they help shape or
evoke. The Dominant Context at any time is a coherent mix of
Qualitative, Conceptual, and Goal Contexts. Our experience at any
time is controlled by numerous mutually consistent contexts. The
reader's experience at this moment is likely to be controlled by
his or her reading of the first part of this book, but also by
much earlier decisions about the difficult issue of conscious
experience, made perhaps many years ago. It is further controlled
by an early life decision to learn to recognize the letter "d"
and distinguish it from "b", "p", and "q", and a later decision
to learn a certain scientific prose style. When these contextual
factors are mutually consistent, they can Œcooperate‹ in gaining
control of the global workspace. If they are mutually
inconsistent --- if the reader years ago decided that
consciousness is utterly hopeless from a scientific point of
view --- the various contextual factors will compete.
4.33 Contexts do not Œcompletely‹ predict conscious
experiences.
In the next chapter we will review the extensive evidence
that completely predictable events fade from consciousness. We
habituate to repeated sounds, we take for granted predictable
thoughts, and we lose consciousness of routine skills. This
implies that context, which constrains many potential degrees of
freedom of a conscious content, does not constrain all of them.
If some input were 100% predictable, we would be habituated to
it, and it would be unconscious. The context of any experience
must leave some degrees of freedom open. In the next chapter we
will develop the argument that consciousness always involves Œa
reduction of uncertainty‹ within a stable dominant context. That
is, the degrees of freedom left by the context are reduced by the
conscious experience, until everything becomes predictable and
unconscious (5.xx).
4.34 Internal consistency and the role of organization in <j <
contexts.
Contexts are organized knowledge structures. This implies
that they are internally consistent; they tend to resist change
when it is inconsistent with context, and resist more strongly
the deeper the inconsistency; there is a tendency to complete
partial input; and when one component changes, another one may
have to compensate. All these phenomena are observable about
contexts. For example, in the Ames room the height of the room is
inconsistent with the person walking in the room --- as he walks
back and forth, his height is perceived to change. As observers
we must revise our contextual assumptions either about human
height or about the room. We first tend to give up our tendency
to see human height as constant. Now we start to toss a ping-pong
ball into the room, and that revision also fails; suddenly we
perceive the room as trapezoidal, and human height become
constant again. There is thus a trade-off between our perception
of height and rectangularity; one or the other changes to
maintain consistency.
The remarkable research tradition on perceptual adaptation
goes back to the 1890s, using distorting goggles, mirrors,
colored glasses, and the like to alter the visual world (Gregory,
1966). Thus the world may be viewed upside down, in mirror image,
with different transformations in the two eyes or even in parts
of each eye, and so on. The auditory realm can be transformed as
well, for example by switching input between the two ears. This
literature demonstrates a truly remarkable ability of the human
nervous system to adapt within a few days to major changes in
sensorimotor context.
Incompatible contexts compete. A scientist cannot
simultaneously presuppose that time is constant and also that it
changes, even when such presupposed knowledge is not currently
conscious. A viewer of the Ames room cannot simultaneously assume
that the visible surfaces are rectangular and that they are
trapezoidal. If there is a conflict between two currently active
contexts, it must give rise to conflicting conscious experiences,
given our previous assumptions (above). These conflicting
conscious experiences presumably serve the cause of
reconciliation between the two contexts. Take jokes as an
example. A typical joke involves the creation of a misleading
context, followed by a rapid restructuring of the information by
a conflicting context. To cite a not very funny but classical
music-hall joke: "Why did the chicken cross the street? To get to
the other side." The question creates a context: we search for an
answer that is not already obvious. The answer repeats something
obvious, thus violating a contextual discourse constraint (x.xx).
The violated context may become briefly conscious --- we may feel
foolish not to have thought of this obvious answer --- and
becoming conscious of this old context may allow us to adapt to
the conflict. The old context may thus become decontextualized,<j <
at least for a moment. Laughter may be one way to adapt to the
conflict.
4.35 Cooperating and competing contexts: Model 2.
Two conscious contents cannot be perceived at the same time.
But two or more contextual constraints can dominate the global
workspace. We previously listed the many simultaneous constraints
that shape our perception of speech (x.xx). Any conscious
experience has many such contextual constraints, perceptual,
conceptual, and goals. We can think of a compatible set of these
contexts as cooperating with each other; incompatible contexts
will compete. Figure x.xx shows a convenient way to diagram both
cooperation and competition. Cooperating contexts are shown in a
hierarchy that dominates the global workspace. Competing contexts
are shown at the same level as the dominant hierarchy, but not
dominating the workspace.
In Chapter 6 we will develop these ideas to claim that
intertwined contexts produce a flow of conscious events that
looks very much like the famous stream of consciousness described
by William James (1890).
Now we can put all the elements together. Figure 4.35 shows
GW Model 2 in which contexts have been added.
--------------------------------
FIG. 4.35, Model 2 Contexts.
SHOW CS INPUT WITHIN UCS CONTEXT
--------------------------------
4.4 Some plausible properties of contexts.
We can think of contexts as topics and themes in the ongoing
conversation of the mind. Conversational topics have a beginning,
a middle, and an end, suggesting that we should explore how
mental topics may be evoked, how they are maintained, modified,
and completed. We will describe the most plausible hypotheses,
based on empirical findings and the logical implications of the
theory so far. Needless to say, these hypotheses require much
more testing.
<j <†
4.41 Accessing and leaving contexts.
The logic of our arguments suggests that to access a new
context we need some distinctive conscious experience. A new
context is by definition a not-completely-predictable event, and
is therefore likely to require consciousness (5.0). A conscious
experience that serves to install a new context may be as simple
as walking into a room, reading a section heading in a book, or
being introduced to a stranger. It may be quite momentary. Or it
may be as complicated, confusing, and upsetting as learning to
perceive a new art form, going through culture shock, or
reluctantly learning some new insight into ourselves. The common
denominator is that a conscious experience give us access to a
new domain of knowledge that is itself largely unconscious, even
though, once accessed it will shape conscious experiences. To
state it more formally:
ŒA major function of conscious experience is to elicit,
modify, and create new contexts --- which in turn set
the stage for later conscious experiences.‹
This point implies that Œtransitions‹ between contexts are
psychologically crucial. In the first moments of a conversation,
many unconscious constraints of the conversation are accessed.
Thus if we were to probe for the parameters of a context, the
ideal time to do so would be at the very start (see 4.xx).
Contexts have a number of other implications, as we see
next.
ŒAn uninterrupted dominant context creates an absorbed state.‹
Occasionally people enter into states that are uninterrupted
for a relatively long time: they may become absorbed in a
fascinating book or creative project, in hypnosis (7.xx), or in a
demanding task, like shadowing speech in one ear. These absorbed
states seem to be controlled uninterruptedly by a coherent
context hierarchy. In absorbed states people resist distraction,
lose track of time, and often report "losing themselves" as well,
suggesting a drop in conscious self-monitoring (9.xx) (Tellegen &
Atkinson, 1974). These are all important features of absorption,
and we will discuss them in some detail later in this book (7.xx;
9.xx).
<j <†
ŒNatural endings and forced exits.‹
The experience of hearing a sentence in a conversation is
constrained by a dominant context hierarchy, including syntactic,
semantic, pragmatic, and discourse components. When the sentence
ends, many of these contexts end as well: there are no syntactic
predictions any more after the end of a sentence; some semantic
predictions may be made about the next sentence; and some
pragmatic purposes will still be unfulfilled. Thus at the end of
a sentence, several contexts come to a natural ending. If the
sentence is, "Nice speaking to you, goodbye," the semantic and
discourse predictions also come to a natural end. Whenever one
context ends, room is made for a previously competing context at
the same level. Thus if we start a conversation to stop feeling
bored, at the end of the conversation, boredom may come back.
This is implied by the notion of a context hierarchy, with
potential competing contexts "lying in wait" for the end of the
dominant context at the same level (4.35).
Natural endings in a context may be difficult to report,
since the context is of course an unconscious structure.
Metacognitive insight into contextual processes may be poor most
of the time, unless the context is disrupted, so that it can
become decontextualized and an object of consciousness in its own
right. In an absorbed state we are generally unaware that we are
absorbed. This state can come to a natural conclusion when its
controlling contexts come to an end, and then we may not even
notice that we were absorbed. But if the state is interrupted, we
may well notice our absorption, apologize for daydreaming, etc.
Metacognitive reports about our controlling contexts are more
likely to be accurate in the second case.
Interruption stops a dominant context hierarchy before its
natural ending. This issue of surprise and disruption is
discussed next.
Œ"Surprise" as a resetting of conscious contents due to
competition between incompatible contexts.‹
Several psychologists have suggested that Œsurprise‹ may reset
conscious contents (e.g. Tomkins, 1962; Underwood, 1982;
Grossberg, 1982; Izard, 1980). It is a plausible idea, which can
be readily interpreted in GW theory. In principle, surprise could
occur with any new conscious content, but if the new content fits
the current Dominant Context, it should not disrupt the context
hierarchy. The next word in this sentence should not
fundamentally jar the reader's conscious experience, because it<j <
fits all the levels of the current context. Truly surprising
events violate deeper layers of context. In the first paragraph
of this chapter we described the experience of going on a short
sailing trip. On stepping back onto dry ground, the scene may
seem to sway for a moment because context predicts a certain
motion in the surrounding world, and this prediction is violated.
This kind of surprise is clearly due to competition between two
incompatible contexts.
Similarly, a graduate student pursuing a PhD may be
surprised to find a book missing in the library. But most of his
or her contextual assumptions and goals remain intact; it is easy
to adapt to the suprising circumstance. On the other hand, if the
student suddenly runs out of money needed to pursue the PhD, this
change of context requires major changes throughout the system of
life goals and local goals.
A violation of one level of context should "reset" the goal
hierarchy, so that the violated context can fragment, but the
higher-level (unviolated) level remains intact. If the book is
not in the library, the goal of finding it may still remain
intact, and we may simply try to find it some other way. If the
book cannot be obtained, we may be able to find the required
information elsewhere. If the information cannot be obtained
anywhere, we may have to change the research project. If the
research project fails, we may have to go into selling life
insureance --- and so on. At each violation of goals the higher
levels remain, and one can rebuild from there to reinstate a
working goal context. Thus "surprise" does not usually imply a
total overthrow of the context hierarchy, only its lower levels.
That is, Œone can think of the context hierarchy as a system that
works to confine change to the lowest possible level‹. High-level
changes are costly, because they propagate throughout the context
hierarchy and require widespread adaptation. Low-level changes
are much to be preferred.
Competing contexts may be involved in episodes of momentary
forgetting and "blanking out". Luborsky (1986) has suggested that
momentary forgetting may occur in psychotherapy when two themes
conflict (Baars, 1986), and Reason describes a number of cases of
forgetfulness and action errors in these terms (Reason, 1983;
1984; Reason & Mycielska, 1982).
Surprise and interruption are central to the issue of
emotion (Mandler, 1975). Surprise creates an Orienting Response,
with major activity in the sympathetic nervous system that
closely resembles emotional upset. Surprise triggers changes in
heart rate, in blood flow patterns, in the smooth musculature of
the digestive tract, and in perspiration, just as emotional upset
does. Personally significant information can be treated in GW
theory as information that triggers deeper levels of the goal
context; naturally this will disrupt lower levels that may be
currently dominant, and it may trigger emotional experienes as
well (Chapter 9). Deeper changes in the goal hierarchy of course<j <
may have consequences that propagate more widely throughout the
context hierarchy.
4.42 Maintaining contexts with conscious reminders.
It is simplistic to think that all contexts are entirely
predictable ove the long term. In any reasonably complex context
there must be points of low predictability, where conscious
involvement is more demanded. That implies that consciousness and
mental effort are required, at least sometimes, to keep track of
current context or goal structure across the gaps in
predictability. We can see this very clearly in the case of
action.
Similarly, it seems that there are Œunder-determined‹ choice
points in the control of action. If we fail to make these choice-
points conscious, errors of action will occur. We can see this in
absent-minded errors. Reason (1984) reports that errors like the
following occur when people fail to pay attention to choice-
points in the flow of action, even though most of the action is
Œroutine‹ and can be considered to be controlled by a single
dominant goal hierarchy. Following are some examples --- errors
occur in the following circumstances. (The examples are
italicized.)
"(1) When a change of goal necessitates a departure from
normal routine.
(ŒI had decided to cut down on my sugar consumption and
wanted to have my cornflakes without it. However, I sprinkled
sugar on my cereal just as I had always done.‹)
(2) When changed circumstances demand some modification of a
pre-established action pattern.
(ŒWe now have two fridges in our kitchen, and yesterday we
moved our food from one to the other. This morning I repeatedly
opened the fridge that used to contain our food.‹)
(3) When we wander into a familiar environment, associated
with habitual activities, in a reduced state of intentionality
(sic). (e.g., in a distracted state - Ed.)
(ŒI went into my room intending to fetch a book. I took off
my rings, looked in the mirror and came out again, forgetting to
pick up the book.‹)
(4) When features of our present environment contain
elements similar to those in more familiar environments. <j <†
(ŒAs I approached the turnstile on my way out of the library,
I pulled out my wallet as if to pay --- although I knew no money
was required.‹)
Apparently consciousness is especially required whenever
there are underdetermined choice-points occur in the flow of
events. Presumably, making these choice-points conscious
facilitates collaborative processing that can resolve the
ambiguities (2.xx). Conversely, a loss of conscious access to
these uncertain choice-points threatens to "derail" the action.
ŒReminding: feedback between context and content.‹
At point of low predictability, unconscious contexts may
need conscious intervention to maintain their stability. It seems
for instance that in listening to speech, limited capacity is
loaded most heavily between sentences and even between the
phrases of a single sentence (Abrams & Bever, 1966). In speaking,
peak load in limited capacity is just before the onset of a
sentence. These findings indicate that there may be a continuous
flow of feedback between conscious content and unconscious
contexts, which over times helps to sustain the dominant context
hierarchy.
4.43 Modifying contexts: coping with the unexpected.
We have previously suggested that the context-hierarchy may
be disrupted by a surprising event, but that generally the
disruption are kept as low-level as possible. What happens with
the disrupted context, however? How is it changed to deal with
the new situation?
Surprise may cause the violated context to be decomposed
into ints component specialized processors. Some of these may
become consciously accessible, so that a group of specialists can
now begin to work together on the job of fixing the fragmented
context. That is consistent with the point made above (4.14,
4.xx) that previously contextual material can become consciously
accessible, at least in part, when the context is disrupted
(decontextualization). Adaptation then becomes equivalent to
cooperative processing between specialists, as discussed in
Chapter 2. When the coalition of specialists from the fragmented<j <
context becomes practiced and proficient again in its new
configuration, it will presumably begin to act as a context
again.
This may be the best way to solve the problem of fixedness
in a single dominant context (4.xx). When context controls
conscious experience so much that a problem cannot be solved, it
may be best to allow the context to be disrupted, so that its
components can be re-examined. This is indeed the principle of
"brain-storming" and other problem-solving techniques. However,
decontextualization no doubt has costs in time and mental
capacity. Fixedness may be especially problematic in panic
situations, where one cannot afford to wait while the context is
disrupted and re-examined (Norman, 1976).
ŒThe assimilation-accomodation dimension.‹
Some challenges to a dominant context are more serious than
others; a mild challenge may be ignored and assimilated, but a
strong challenge must be treated more seriously (5.0). The
following example is in the conceptual realm, but the same point
applies to all kinds of contextual knowledge. Consider the
following set of questions from Eriksen and Mattson (1981).
(1) How many animals of each kind did Moses bring on the
Ark?
(2) In the Biblical story, what was Joshua swallowed by?
(3) What is the nationality of Thomas Edison, inventor of
the telephone?
Some readers no doubt noticed that each question contains a
flaw, but most experimental subjects did not notice any problem.
They simply answer "two" to question (1). But when they are
asked "who built the Ark in the Bible?" they will correctly
answer "Noah," showing that they do know the correct answer.
Further, their immediate memory for the sentence is quite good,
because that they can repeat each question accurately, without
spotting the incorrect name. But somehow they do not Œbring their
knowledge of the correct name to bear on the interpretation of
the question.‹
Now consider what happens when the flawed name is changed to
one which violates context much more severely:
(1) How many animals of each kind did Nixon bring on the <j <
Ark?
(2) In the Biblical story, what was Jeffrey swallowed by?
(3) What is the nationality of Benjamin Franklin, inventor
of the telephone?
Now no one is fooled (Eriksen & Mattson, 1981). This
"semantic illusion" illustrates our general claims about the
context of experience very well. When the context is weakly
violated, people usually do not even perceive the error; when it
is strongly violated, they become conscious of it, so that what
was previously context becomes conscious content (4.xx).
We find similar phenomena in the well-known Proofreader
Effect, the general finding that spelling errors in page proofs
are difficult to detect because the mind "fills in" the correct
information. Here, too, we expect to find a difference depending
upon the degree of error: perhaps spelling errors that would
change the pronunciation of a word are more egregious than those
that do not, so that they are more likely to become conscious.
Similarly, errors and dysfluencies in speech are surprisingly
difficult for people to detect (MacKay, 1981). If we are
listening for the intended meaning rather than for errors, as we
normally do, minor errors are rarely noticed.
Input that violates minimally is not consciously perceived;
input that violates moderately and that can be framed in another
context is consciously perceived; and input that violates context
totally and utterly is not consciously perceived in that
particular framework. It may be dissociated --- it is treated as
something else, it may acquire its own context.
4.5 Implications for empirical testing.
4.51 Related concepts.
Context effects are so powerful and pervasive that every
psychologist has surely observed them. But they are seldom stated
in terms of the influence of unconscious events on conscious
ones. Even current cognitive work on context tends to evade this
issue, which is central for us. However, the existing research
literature is easy to reinterpret in these terms. Perhaps the
best worked-out theory of context is Adaptation Level (AL)
Theory, originally developed by Helson (e.g., 1964). AL theory
predicts that the perceived amount of a perceptual or conceptual<j <
event depends on previous exposures to a range of similar events.
Thus one's judgment of the severity of a criminal offense depends
on previous exposures to a range of criminal offenses that differ
in severity. After exposure to mass murder, mere assault seems
less serious. The same pattern of adaptation by contextual
manipulation applies to perceptual evaluations and to concepts
like happiness, cost, and the like. There is also a considerable
linguistic literature on topics like focus and presupposition,
given vs. new information, topic vs. comment, etc. (Clark &
Clark, 1977) --- all of which correspond to context and content
respectively. In the study of expert systems there is now an
acute awareness of the presupposed and automatic nature of expert
knowledge, compared to the same knowledge when it has been newly
acquired (Anderson, 1983). On the theoretical side, Minsky's work
on "frames" and Bransford's research on "activated knowledge"
seem closest to our claims about contexts (Minsky, 1975;
Bransford, 1979).
4.52 Some testable predictions from Model 2.
The evidence for contexts discussed above (4.1) can also be
used to test our predictions. Priming effects, fixedness, top-
down influences, and reports of violations of contexts may all be
useful. There is currently a very interesting empirical
literature on priming effects for assessing a variety of
cognitive processes, which may be adapted to testing questions
about context.
ŒContextual transitions as priming events.‹
This chapters suggests that in contextual transitions,
things become momentarily conscious that are normally unconcious
in the midst of a context. The example of gaining one's sea legs
is illustrative. People should rate the swaying of a boat more
highly if they have just stepped onto it, than later on.
Similarly, we may quickly be conscious of our attitudes toward a
friend at the very beginning of a meeting; a few minutes later
those attitudes may have become inaccessible, at least until we
take our leave. If we could probe someone at the beginning of a
conversation, we should find that thoughts of friendship or
hostility are more accessible than later on.
If a change in context becomes at least partly conscious, a
switch of interpretation of an ambiguous stimulus should load
limited capacity. If we measure limited capacity by a dual-task
technique, for example (1.xx), we should find that when we<j <
switch between two views of a Necker cube, two interpretations of
a joke or an ambiguous sentence, or a change of mind about
someone's motivation --- all these contextual transitions should
load limited capacity. Perhaps measures of surprise, like the
Orienting Response, will also show measurable effects, though
this may depend upon the "depth" of contextual violation involved
(4.xx).
ŒContexts may have many simultaneous levels.‹
Context may be a many-leveled thing. If two contextual
systems simultaneously constrain conscious experience and access,
then a conscious experience that fits both of them should be
favored. We have already described the experimental use of Œdouble
entendres‹, linguistic ambiguities designed to fit two different
contexts, and the fact that they are indeed chosen more often if
both contexts are activated. Similarly, experimentally induced
slips of the tongue, which tap into two simultaneous contexts are
more likely to occur than slips that reflect only one context
(7.xx). These are purely laboratory demonstrations, but Spence,
Scarborough & Ginsberg (1978) have shown that in spontaneous
speech, terminal cancer patients produce more Œdouble entendres‹
related to death and disease than controls. Similar studies
could be carried out with any dominant mental set.
ŒBlindness to conceptual presuppositions‹
Although there is considerable work on fixedness in problem
solving, there is very little research exploring the apparent
difficulty of accessing presupposed knowledge. Yet the
naturalistic evidence in favor of "presuppositional blindness"
is very strong --- everyone must surely have encountered an
inability to remember one's own basic beliefs, even when those
beliefs continue to guide one's actions. Many people, scholars
especially, routinely deal with students and skeptics who demand
an explicit account of those beliefs, and naturally, if we are in
the habit of doing this kind of "justificational" argument,
accessing routine beliefs becomes easire. But this relative ease
of access is misleading: it is not representative of the great
bulk of mankind, which does not need to justify its
presuppositions very often.
It seems that presuppositions blindness involves not just a
problem in accessing overarching beliefs, but an inability to see
and "hold on to" the negation of some belief. This is especially
obvious in the cases of scientific change mentioned above (4.xx),
but it may also account for resistance to attitude change as
described in numerous social psychological studies (Abelson et<j <
al, 1968).
4.53 Some questions Model 2 does not answer.
So far, we have dealt with contexts without paying much
attention to the ways in which they develop and are changed. We
have not investigated them over time. In the next chapter we will
try to correct this deficiency. We will pursue the claim that
conscious experiences, when they are adapted to, result in new
contexts, which, in turn, serve to constrain later conscious
experiences. Thus we are always defining and redefining our
reality, but "getting used to" new experiences. The resulting
perspective has strong implications for learning.
We have not addressed the issues of goals, intentions, and
voluntary control so far. A complete theory must deal with these
fundamental topics, and we develop an approach to them in
chapters 6 and 7.
4.6 Summary and a look ahead.
We have explored the pervasive influence of unconscious
systems that act to evoke, select, and shape conscious
experience. The evidence for such effects is very strong. Indeed,
there is no conscious content without context. Psychologically we
are always ensconced in the midst of a multitude of ongoing
unconscious systems, which shape, and define our experience. Some
of these unconscious systems have been with us from childhood,
while others may have been evoked in the last few minutes.
In GW theory we can view contexts as collections of
specialists that are committed to a certain way of processing
input, and which can come to dominate the global workspace at
least for some time. We can specify different contexts for
perception and imagery (where they help shape qualitative
experiences), and in conceptual thought, goal directed
actitivities and the like (where contexts serve to access
conscious experiences). The results can be modeled with little
difficulty in an extended version of GW theory. It seems that one
main function of consciousness is to evoke contexts that will
then shape later conscious experiences.
Experimental psychologists often seem to feel that context
effects are to be controlled and eliminated from an experiment if
at all possible. This, we would argue, is a mistake. One can
indeed suggest that some of the most serious conceptual errors in<j <
psychological history --- errors that misled researchers for
decades --- began with naive attempts to remove phenomena from
their natural contexts. We would argue that that context effects
are impossible to eliminate, and that we should not wish to
eliminate them totally, but only to study them. There is no zero
point in the flow of contexts. They are not incidental phenomena
that confound our careful experiments: they are Œquintessential‹ in
psychology. There is no experience without context.
ŒFootnotes‹
1. The notion of a context developed in this chapter and the
next owes much to many fruitful discussions with Michael A.
Wapner.
2. We may sometimes want to treat "context" not as a thing
but as a relationship. The assumption made by the visual system
that light comes from above may be said to be "contextual with
respect to" the perception of concavity of moon craters (Rock,
1983); likewise, an implicit moral framework may be "contextual
with respect to" one's feelings of self-esteem. There is no need
to become fixated on the question whether context is a thing or a
relationship. In either case, contextual information is
something unconscious and stable that profoundly shapes whatever
becomes conscious.