What is the Unconscious?

Certain mental processes can affect our consciousness and
behavior. We cannot, or only with great difficulty, make verbal accounts
of these processes, or in other ways bring these processes into our
awareness. A common name for these processes is The Unconscious
(Kihlstrom, 1990, p 446 og 448; Lazarus, 1991, p 361).

The original project of empirical psychology was the study of those
mental processes that are accessible to our conscious awareness. Wundt,
Titschener, and other structuralists who helped initiate psychological
research assumed that the mind in general is capable of observing its
own innter workings (Kihlstrom, 1990, p 445).

However, there also existed a strong philosophical tradition that
focused on mental processes not available to the conscious.
Arthur Schopenhauer (Schopenhauer, 1819, cited in Hogan, 1985,
pp 48-49) argued that humans are basically irrational, and that they
are pushed forward by inner forces that they are only vaguely aware of,
the stronges of these being the desire for reproduction. Nietzsche
(ibid.) argues that any emotion, attitude, act or expressed
virtue is rooted in self-deception and is founded on an unconscious lie
- they are sublimated expressions of sexuality and aggression.

Freud argued that our conscious mental life is determined by
uconsciuos ideas plus impulses and emotions, together with defence
mechanisms that are unconsciously used against these. Helmholtz, on the
other hand, argued that conscious perceptions are the end result of
unconscious inferences based on the individual's knowledge of the world
and memories of past experiences; and that e.g. visual perception
partially must be based on rational but unconscious
inferences that express themselves to the conscious. An example of this
is depth vision (Johnson-Laird, 1988, p 15, p 354). Ever since, the
study of unconscious has been divided into two camps. The
dynamic and the cognitive unconscious have been
studied independently and without reference to each other (Boston, 1986,
cited in Kihlstrom, 1990, p 446).

The question that has intrigued researchers and philosophers ever
since is this: How do thoughts and impulses become and remain
unconscious? I will concentraite on explanations from cognitive and
dynamic traditions.

In the following, I will concentrate on the dynamic unconscious as it
is described in e.g. Pervin (1990) and Gullestad (1992), and the
cognitive unconscious as it is described in e.g. Kihlstrom (1990) and in
Lazarus (1991).

I will discuss the unconscius from the perspective of recent theory
and research. In particular, I would like to discuss

1. Which tasks are unconscious?
2. What is the relationship between the unconscious and emotions?
3. What is the relationship between the cognitive and the dynamic unconscious?

Background

The Unconscious according to Freud

Freud viewed the human mental apparatus as an energy system where
energy flows, is side-tracked, or is halted. This energy is limited,
meaning that if much energy is used for one thing, correspondinlgly less
is left for other things. The goal of all behavior is pleasure,
in the shape of reduced tension or emission of energy (Pervin, 1989,
p. 77).

Mental life can be described to the extent to which we are aware of
phenomena. The conscius refers to what we are aware of at any
given time, the pre-conscius refers to what we are not
currently aware of but can be made aware of if we direct our attention
towards it, and the unconscious refers to those phenomena that
we are not aware of and that we cannot be made aware of except under
very special circumstances (Pervin, 1989, p 80).

The unconscious is alogical and does not obey the normal constraints
of time and space. It is not noticed during the main course of our
normal lifes, but it can be made more lucid during dreams and in the
ideas of psychotics.

Freud found evidence of the unconscious during his clinical
work. Patiens were, during the course of their therapeutic process,
aware of memories and desires of which they were previously unaware.
These discoveries were usually associated with painful emotions. This
gave Freud the idea that the unconscious contains memories and desires
that are not part of our conscious awareness, but rather actively
hidden in the unconscious. (Pervin, 1989, p 83)

The theory of repression is the corner-stone on which the whole
structure of psycho-analysis rests (Freud, 1914, cited in Gullestad,
1992, p 29).

The concept, repression, has been further elaborated since
Freud's days. It is still assumed that repression is a defence of the
ego against inner dangers, analogous to escape as a defence against
external dangers (Gullestad, 1992, p 815). By and by, Freud extended
the concept of reprssion to include more than inner defence. In Freud's
later structural theories, anxiety aquires a key role as a
motif for repression: Anxiety implies an anticipation of a dangerous
situation, and is defined by helplessness (Freud, 1923, 1926,
cited in Gullestad, 1992, p 816).

The principles behind repression remain the same: The individual
refuses to recognize something that evokes an unacceptably high degree
of anxiety. Therefore, the anxiety-provoking idea becomes repressed, in
that the individual hides it for him or herself (Gullestad, 1992,
p. 816). According to Sartre, this implies a paradox:

In order to perform its activities, the censor must know that
which shall be repressed. The censor must decide whether
something can pass or be repressed. How can the censor at the same time
both know and not know of something? How can it decide not to know
something that it knows? (Sartre, 1943, cited in Gullestad, 1992,
p. 818).

If repression is assumed to be motivated, i.e. not as an
expression of causal processes this ceases to be a paradox, argues
Gullestad (1992, p 818). One does not decide to repress the
unallowable. The word decide must be viewed as metaphoric. But
how can the repression in itself be unconscious? Gillet, 1987, cited in
Gullestad (1992, p 818) resolves this by postulating two
unconsciouses: in addition to the dynamic unconscous, there
exists a descriptive unconscious that is unconsius in a
descriptive sense, which cannot be experienced and never be made
available to conscious awareness.

Such as this is described in Gullestad (1992), this appears to me to
be a bit too ad hoc: It stands to reason that cogntive,
biological processes cannot ever be made available to the conscius, but
why should the mechanisms of repression be part of them, unless they can
be shown to have a biological correlate?

In the following, I will present theory and research that expands the
view of which tasks the unconscous may have, and where it is not
unlikely that defence mechanisms may have their place, together with
other non-conscious cognitive processes.

Out of all of the academic psychologists of the 19th century, William
James was possibly the most ardent advocate of unconsious mental life
(Kihlstrom, 1990, p 448). James argued that mental states can be
unconscious in at least two ways: First, a mental event can be actively
left out of awareness; second, he argued for a division of consciousness
into primary, secondary etc. consciousnesses. Only one of
these is available for phenomenological awareness at any one given time.
He preferd to talk about co-conscious or subconscious processes rather
than unconscious processes to avoid misunderstandings.

Broadbent's Filter Model

It is a common experience that we can only concentrate on a
little of all that we may concentrate on. We do not
notice everything we see at the time at which we see it. But can that
of which we are not aware, nevertheless affect us?

An early idea is that those stimuli that awareness is not directed
towards is filtered away at an early stage of our awareness system
(Broadbent, 1958, cited in Atkinson et al., 1987, p 198). It
has later been demonstrated that quite a bit of processing must take
place prior to selection because a message may be accepted or denied on
basis of its significance. Our perceptual system stores material that
are not in our awareness for some time and processes it to some extent.

One's own name can be picked up, even it it is uttered quite softly
in a conversation one is not part of (the cocktail party
phenomenon). This would not happen if the entire conversation is
filtered away at an early stage in our perceptual system (Atkinson
et al., 1987, p 111).

The Hidden Observer

Hilgard (1982, cited in Atkinson et al., 1987, p 40)
describes a hypnosis-situation where the hypnotized subject was given
the suggestion that he was deaf as long as Hilgard's hand was on his
shoulder. This individual was also blind, and with a hypnotically
induced deafness the subject was pretty much cut away from his
surroundings. Hilgard was asked by the audience if the subject was as
unresponsive as he seemed. Hilgard then asked the subject to raise his
finger if any part of him was capable of hearing. To everybody's
surprise, including Hilgard and the subject, the finger raised.

This is part of the evidence of a hidden observer, a mental
structure that monitors eevrything that happens, including things that
occur outside the hypnotized subject's awareness.

I take this to imply that there exist processes that we are not aware
of, but can nevertheless affect our behavior. These processes can be
mental, they are at least capable of processing meaningful material.
But to how great an extent can unconscious processes contribute to our
behavior? Recent theory and research suggest that it can be quite a
lot.

Models from Cogntivie Science

The expressed project of Artificial Intelligence (AI), an
important part of Cognitive Scicen, is to construct a machine
that, based on theories of how humans think, can think and possess a
conscios (Penrose, 1989, p 3; Searle, 1987, p 18; Johnson-Laird, 1988,
p. 26). The corner stone of AI is the theory of a relatively simple
mathematical model that can compute any computable mathematical
procedure, by manipulating simple symbols according to certain
rules.

Why is a machine that can achieve consciousness of interest
in a discussion of the unconscious? Because it has been
suggested that the most recent AI research is not so much about
consciousness as about unconsciousness.

Connectionist-systems (hereinafter PDP-networks, from Parallel
Distributed Processing has been suggested as an alternative
approach for testing hypotheses about cognitive processes. They work,
according to Clarke (1987, p 4) metaphorically like the open market:
Global patterns of offers and requests are established according to
local interactions between sale and purchase. Higher-level knowledge of
current demands is dispersed between the buyers, and higher-level
control of goods between the suppliers.

In The Emperors's New Mind (1989), the British phycisist Roger
Penrose argues that one of the main problems with AI's project is that
the present knowledge of the laws of physics is not sufficient to
understand how to create a machine that thinks (p 16). In the section
Parallel computers and the oneness of consciousness he writes

Many people appear to be of the opinion that the development of parallel
computers hold the key to building a machine with the capabilities of a
human brain. [ ] A parallel computer, as opposed to a serial one, has a
very great number of separate calculations carried out independently,
and the results of these largely autonomous operations are only
intermittently combined together to give contributions to the overall
calculation. The motivation for this type of computer architecture comes
largely from an attempt to imitate the operations of the nervous system,
since different parts of the brain indeed seem to carry out separate and
independent calculational functions (e.g. with the processing of visual
information in the visual cortex) (Penrose 1989, s. 389)

One of the characteristics of conscious thought is that it is
concentrated on one thing at a time. Thinking of several things at
once, at the same time, independend and undisturbed by each others, is
impossible, says Penrose (1989, p 399), and in that respect the parallel
computer serves better as a metaphor of the unconscious.

The Boltmann-machine is one illustrative example of a PDP-network
(Ackley et al., cited in Clarke, 1987, p 4). Is is composed of
many simple units that work in parallel and are connected with their
neighbors by way of links where the stream of information can proceed
both ways. Clark (ibid.) suggests that every single unit may
be viewed as a primitive hypothesis of one goal area or another. Two
units that represent contradictory hypotheses can be viewed as a
connection that weighs negatively. Hypotheses that support each other
are positive connections, etc. These units will affect and be affected
by other units again. In the end all the units together will be
adjusted towords a common decision. This process is called
relaxation (Clark, 1987, p 4).

The PDP-netwoek has two advantages. First, PDP-networks are fast
because many processes do not need to wait for each other to finish to
the same extent as a serial machine. Second, PDP-networks are far more
robust. If a part of a well-constructed PDP network is destroyed, the
network does not stop working, but it functions poorer, perhaps in the
same way that the brain is reduced by age. This phenomenon is known as
Graceful Degradation (Clark, 1987, p 6).

Broadbent (1985, cited in Clark, 1987, p. 9) argues that PDP-networks
are perhaps just one of several useful ways of describing cognitive
processes, and argues that PDP-networks do not have psychological
relevans as such. McClelland & Rumelhart (ibid.) oppose this
position and argue that the level at which processing occurs is
theoretically relevant.

From an evolutionary perspective, Clark (1987, p 9) argues, the brain
may have started as a PDP network, with all of the advantages this has:
Conclusions are drawn rapidly, and the inderlying machinery is
relatively robust and provides a strong foundation for further
development. However, the sequential Turing machine also has
advantages, e.g. they are more accurate: it is easier to describe
formal reasoning in a Turing machine. Perhaps, concludes Clark (1987, p
12), the brain is a PDP network that uses Turing machines for the more
specialized, higher-order processes. This depends on whether or not
PDP-networks are able to construct Turing machines.

Unconscious cognitive processing

A large part of mental activity is unconscious in the sense that it
is never under any circumstances available for phenomenological
knowledge. Some unconscious knowledge seems to be inborn, other
cognitive and motoric knowledge that is aquired by experience seems to
be made routine by rrepeated exercise. This process may be described as
Compilation of Knowledge, indicating that the format in which
knowledge has been represented, has been changed. Thus, both inbornd
and aquired knwoledge may be unconscious (Anderson, 1982, Nisbett
Wilson, 1977, cited in Kihlstrom 1990, s. 448-449).

What is needed for behavior to become automated, as opposed to
intended? Most importantly, that it requires little or no cognitive
capacity, so that it can be performed at the same time as intended
actions. An example of advanced cognitive processing is provided in
Mærk Verden:

Modern football is characterized by many and trained patterns. The real
geniuses such as Pelé, Cruyff, Netzer, Maradona, and Landrup break the
patterns all the time. That's why they are so good. (Nørretranders,
1991, p 312, my translation)

If the football players had to think prior to each decision they
wouldn't have had the time to play at the level they do. The decisions
are made quickly, accurately, and unconsciously.

It has been argued (Pierce Jastrow, 1884, cited in Kihlstrom, 1990, p
450) that an event cannot be analyzed for its meaning unless it has been
identified and viewed consciously. It has been shown that this is not
necessarily the case. Marcel (1983a, 1983b, referert i Kihlstrom, 1990,
s. 450) employed a lexical decision task where one stimulus-word (prime)
was followed by another stimulus-word (target). Subjects were asked to
decide whether the target word had meaning. The judgements were more
accurate in the cases where both target and prime had meanings. Marcell
superseded his primes with masking stimuli, so that the subjects could
not reliably detect the primes. Even so the effect of the primes was to
facilitate the lexical decision task. Since semantic priming requires
some semantic processing, it seems that analyses of meaning can take
place for stimuli kept outside of conscious knowledge.

It appears that the effect of semantic priming can reflect aquisition
of novel knowledge, or just activation of existing knowledge. When
subjects study lists of associative pairs, such as
sour/grapes and small/potatoes, and
are then asked to state the first word they can think of when cue is
sour or small, subjects display a priming of the target response
independent of the degree to which they are able of recalling the
word-pars with which they are already presented. Since such phrases as
sour grapes or small potatoes are common English
expressions, priming seems to activate knowledge aldready stored in
semantic memory.

It appears,, infers Kihlstrom (1990, p 451) that implicit perception is
subject to the same constraints, i.e. events in the relevant
surroundings can directly affect mental functioning outside of conscious
knowledge, but only if the event activates existing knowledge or if the
individual affords active attention towoards that part of the
stimulus-field in which the event is taking place. These assumtions
remain to be tested (ibid.).

There is reason, then, to assume implicit or non-conscious memory and
perception (ibid.). However, there also exists evidence for
implicit thinking and learning. In one experiment on
implicit learning, subject's ability to use rules aquired by experience,
but without knowning the rules themselves, was investigated. Implicit
learing has already been demonstrated, in that humans when aquiring
language learn to differentiated between grammatical and ungrammatical
sentences (Kihlstrom, 1990, p 452). Reber (1976, 1989, cited in
Kihlstrom, 1990, p 452) has attempted to model this process in the
laboratory by developing artificial grammars. The rules of these
grammars constrain the construction of well-formed letter
sequences.

In Reber's procedure, subjects learned by heart e.g. 20 letter strings,
put together according to a set of about 20 rules. Examples of
grammatically correct strings are PVPXVPS or PTTTVPS. Subjects are then
shown a number of new strings, some of these are grammatically correct,
such as PTTTTVPS, and some are not, such as PTVPXVSP. Even though
subject are not capable of fomulating the underlying grammatical rules
that they have inferred from the strings they have already seen, they
are capable of telling grammatical from ungrammatical strings on a level
better than chance. Kihlstrom (1990, p 453) argues, citing even more
experiments, that it is reasonabel to assume that these experiments show
aquisition of new knowledge even when there is no conscious intention to
learn, and no conscious knowledge of what has been learnt.

Kihlstrom (1990, p 456) postulates two general types of
conculsions. First, the conscious cannot be identified by any
specific perceptual or cognitive functions - alle these functions can
occur without phenomenological knowledge. Consciousness may, instead,
be viewed as a quality of attention that can follow all of these
functions. Conscious knowledge may have definite consequences for
psychological functioning: they are necessary for willed control; for
communicating own mental states to others; and for instructing others.
But consciousness is not necessary for many forms of conscious
psychological functioning. Second, Kihlstrom (1990, p 457) assumes a
temporary classification of non-conscious mental structures and
processes that have a foundational character for the areas of control
for the cognitive unconscious.

Within procedural knowledge, there exists a number of komplex
processes that are properly unconscious (Kihlstrom 1990, p 457). This
is to say that they are under no circumstances available to the
conscious. They operate on declarative knowledge with no conscious
intent and no conscious awareness, either because they have become
automated or because they are inborn. These constitute the unconscious
proper.

In principle, declarative knowledge is available to phenomenological
knowledge, and may be brought to awareness by introspection or
retrospection. However, it should now be clear that procedural
knowledge may interreact with, and take advantage of declarative
knowledge which itself is not available to conscious awareness.
Implicit knowledge and menory indicate that there exists a category of
pre-conscious declarative knowledge structures. As opposed to
properly unconscious procedural knowledge, this knowledge is available
to the conscious under normal circumstances, but it does not itself
cross the threshold for representation in working memory and thus be
made available for conscious knowledge. These representations, that make
the foundations fro implicit perception and memory, are at the border of
consciousness. If situations change, these may in principle be made
available to the conscious.

Hypnosis and related dissociative phenomena seem to be examples of a
kind of subconscious declarative knowledge. This declarative
knowledge can be used by consious knowledge, and it covers cognitive
processes with such a high degree of activation that this should be
sufficient for representation in working memory. Nevertheless, it
appears that this knowledge as such is not available to conscious
knowledge (Kihlstrom, 1990, p 457).

Emotions and the unconscious

Some researchers view all mental phenomena as identical to
consciousness -- everything else is newrophysiology (Johnson-Laird 1988,
p 354). Freud assumed an additional unconscious hideaway for primitive
instinctive drives and suchlike, that are too anxiety-provoking to be
allowed into the conscious (ibid.). Helmholtz' view is that the
unconscious, additionally, harbor benevolent processes that form the
base for thinking and perception. The existence of such processes
indicate that mental processes can work in parallell (ibid.).

Parallell processing may happen in many ways, posits Johnson-Laird
(1998, p 355). It cannot compute what a simple sequential machine
cannot compute, but it can do so much faster, and the risk that damage
to the system will render it unusable is lower. Several researchers of
Cognitive Science claim, writes Johnson-Laird, that the brain
could not have evolved if it was not divided into separate modules.
However there may be conflict between the modules if modules within the
same domain give contradictory messages. One reasonable way to solve
this would be to have classes of low-level processes being supervised at
a higeher level, in a hierarchy where processes have increasingly larger
areas of responsibility. Processes at a higher level can then decide in
lieu of condradictory lower-level processes, in order to ensure that the
system does not go into a clinch.

Plain consciousness, posits, Johnson-Laird (1988, p 356), as in pure
experience of pain, may have had its origin in the developmen of
higher-level monitors from the network of parallell processes. This
operating system on top of the hierarchy decides the goals for
the lower-leves processes and monitors their efforts. Since it is on
top, its instructions can be specified symbolically, as e.g. rise and
walk!. These instructions will be interpreted and passed downwards
as increasingly more fine-grained instructions down to the contraction
of muscular celles (p 356).

The experience of reality is a triumph of the architechture of the
mind, maintains Johson-Laird (ibid.) The operating system has no access
to the processes on which it is based. There are sound evolutionary
reasons for this, he argues: If it were possible to reflect upon the
entire perceptual process, it would have to be much slower because it
would not be possible to rely on parallell processing. It would also be
possible to doubt one's own senses, something that could have dramatic
consequences in a dangerous situation.

This theory implies that there exists a divide between the conscious,
and unconscious processes lower down in the hierarchy. Such
dissociations may actually exist in human beings (p 357).

The phenomenon, blindsight is one striking disocciation. After
damage to the striate cortex of the occipital lobe (a part of the brain
that is assosiated with sight) (Weiskrantz, 1980,
Weiskrantz et al., 1974, cited in Kihlstrom, 1990, p 453) some patients
claim that a relatively large area of their vieual range has become
blind. If patients, however, are asked to guess the location of a spot
of light within their blind range, they are remarkably good at it.

Another form of dissociation (p 358) is self-conception.
The most remarkable example of this is hysterical paralyses, which are
not due to damage to the central nervous system, but
unconscious motives ¹. Lord
Adrian (Adrian Yealland, 1917, cited in p 358) cured neurotic paralyses
that were results of Shell shock during WWI. He assumed that the
patients were not simulating, but that they were convinced that
they were actually paralyzed. Convincing them of anything else was
impossible. He was confined to trick those of the patient'
unconscious processes that were responsible for the condition,
and treated the paralysis as if it were real. By using a simple
procedure that would not have been able to cure a real paralysis, he
cured 90 percent of his cases. Johnson-Laird (p 358) argues that
cognitive therapy uses a similar approach: induce a change of
unconscious processes.

If one views dissociation from a cognitive perspective, a very different
view emerges than when viewed from a psychoanalytical perspective
(Hilgard, 1977; Kihlstrom, 1984, cited in Kihlstrom, 1990, p 455). Non-conscious mental content does not need to be limited to
primitive sexual or aggressive ideas and impulses, and need not be
irrational, a-logical, or in any way radically different from conscious
cognitive processes. The only real difference is that they are not
available to consciousness (Kihlstrom, 1990, p 455). The
separation from the conscious needs not be motivated by defence, or have
reduction of anxiety as a result. On the contrary, the dissociation
from the conscious may be result of certain psychological operations
(Kihlstrom, 1990, p 455).

Thus, nonconscious mental processes are not restricted to automatized
procedureal knowledge, and nonconscious mental contents are not limited
to unattended or degraded percepts and memories. These differences
suggest that dissociative processes deserve more attention from both
cognitive and clinical psychologists than they have received in the
recent past. (Kihlstrom 1990,p 455)

Reflection upon the self, claims Johnson-Laird (1988, p 361), may happen
by consciously producing mental models of oneself. Just like a
robot can avoid falling into pits or run into things, because it has an
internal representation of its environment as a guide, humans can form
internal representations of their environment, including themselves
(p 360). This assumes that the framework of reflections and intentions
that we use to reflect on our behavior, and to explain it,
is no epiphenomenon, but plays a causal role in our thoughts and
actions. This theory gives meaning to a lot of phenomena (p 362).

If one has succeeded in solving a lot of exersices of the same type, and
then suddenly fails, one may ask: What do I do when I solve
these exercises? The answer is based on the ability to view oneself
- to step up one level, as it were, so as to view oneself as a
spectator would. The anser kan help re-formulating how to carry on, at
the same level as the actual act takes place (Johnson-Laird, 1998,
p. 362).

Not only self-spectating is possible. It is also possible to step up
one more level to consider one's observations of oneself, to observe
that observation, etc. Free will, Johnson-Laird (1998, p. 366),
consists of the ability to choose at which level one wants to observe
oneself, and to act in accordance with that particular level.

Consciousness, suggests Johnson-Larid (1998, p. 367), is a feature of a
certain class of computational procedures. The result of these
procedures may, in principle, be taken from serial procedures that are
incapable of giving consciousness. Thre is, then, no way to observe
conscious behavior to ensure that it is conscious. Nevertheless, in
everyday life it is possible to determine whether someone is conscious
or not. Since consciousness is the highest level of reflection, it is
possible to determine the extent at which an individual is capable of
self-reflection. Individuals who can report that they chose a certain
action because they weighted alternatives and in the end decided upon
that action because that action was considered the best, can be counted
for responsible, consciously acting individuals who possess free will.
Patients who suffer from brain damage or mental disorders may be thougt
of as not having access to models of themselves or a reflecting
procedure, and will not be able to take part in a self-reflecting
discussing Stress, worry and other factors contribute to human's
tendency to act wrong.