Caroline: Introduction II

I will begin this discussion of dreaming by quoting Joseph Weiss, who believes that "a person, in his dreams as in his conscious waking life, thinks about his reality and attempts to adapt to it. A person's dreams are products of normal (albeit unconscious), thoughts, and they express his attempts at adaptation. He produces dreams in an effort to deal with current concerns that he has not resolved by waking thoughts, either because these concerns are too overwhelming, because he is hampered in his thinking about them by his pathogenic beliefs, or because he has not had time to think about them. A person may sometimes reveal more self-knowledge and may see things more clearly in his dreams than in his waking life. In his dreams a person may assess situations and develop plans for dealing with them much as he does in waking life. He may alert himself to a problem that he has overlooked, make a resolution, remind himself of a new insight, console himself for a loss, or reprimand himself for a misdeed. He may prepare for an upcoming task by encouraging himself; or he may bring forth repressed traumatic experiences so as to make himself aware of the traumas and to master the affects connected with them; or he may tell himself more clearly than in waking life how he feels about someone who is close to him; and so forth. In a sense a person, by producing a dream, sends a message to himself."

This is exactly my view of dreaming. Much of my effort over the last twenty or so years has been to hold to this viewpoint by working out some logical problems with it. One of these is simply, if we are the source and the recipient of the messages we send ourselves via dreams, why do we need to dream? What does a dream tell us that we don't already know? If we assume that a dream does tell us something important and new, it follows that we must be attending to the meaning of the dream as it unfolds. But how can this be, if we are creating the dream on a moment-to-moment basis at the same time? It seems credible that we could do one or the other, but not both at the same time.

Weiss's theory is about the dream we can sometimes remember when we wake up. Since it is more probable that we remember a rapid-eye-movement (REM) dream than a non-REM dream, Weiss's theory would seem to be about REM dreams. But we have 4-6 of those dreams every night. If we send messages to ourselves in our dreams, why do we need to do it so many times? And if dreams are so rational, why do they seem so crazy?

I will start with the simplest problem. How is it that we can make up a dream and attend to it at the same time? We can't. Sleep almost invariably begins with non-REM episodes, activities which serve to prepare the dream, I believe. Each REM dream has been constructed beforehand, so that we can devote our attention to the meaning of our REM dreams as we dream them.

What is different about REM dreams? Both REM and non-REM dream thoughts are unconscious, but REM dream thoughts are experienced in a literal sense, I believe, whereas non-REM dream thoughts are not.

It is possible to attempt to use this distinction in answering the questions I raised in terms of ad hoc rules of thumb, which is what I presented in my previous posts on dreams. I believe now, however, that it may be possible to do better by delving into the nascent field of cognitive neuroscience. The basic metaphor here is that of the neural network.

There is a danger in taking this metaphor too literally. The concept of neural networks was invented by physical scientists who had no real idea about how the human brain works, so there are potential problems with using this concept to explain brain activity. Happily, I can answer the questions I raised, at least to my own satisfaction, without resorting to any detailed neural-net model. About all I need are the language of neural nets and a few empirical facts about how neurons function.

In the 1860s the British neurologist John Hughlings Jackson wrote that a specific mental ability is not produced by one distinct group of localized brain cells, but rather arises at several levels of organization, beginning with a low (spinal or brain stem) level, a motor or sensory level, and then at a "high" frontal level. The more complex the activity, Jackson believed, the more brain processes are recruited--more in reading a word than a letter; more in reading a sentence than a word; and more in reading a story than a sentence. The more complex the activity, the more combinations of different processes are needed to carry it out.

Jackson's view lives on in modern cognitive neuroscience, which postulates that simple functions are carried out by relatively few neurons acting in concert either as a single network or as a small number of networks to produce a relatively simple internal mental outcome or overt action. The resultant outputs of the various neuronal groups may act as a component of a skilled performance, for example. The entire skilled performance requires more neuronal groups, and with increases in skill, even more groups become engaged in integrated and coordinated ways to better achieve the goal of the activity. The integration of several skills involves enlisting the concerted activity of even more neural networks. Presumably in the psychological realm, the brain is acting pretty much as a whole.

People who theorize in terms of neural nets imagine neurons connected to other neurons to form networks, each of which receives signals from sensory inputs and/or other networks and responds to these inputs by generating a set of output signals. What is imagined is that initially for a given input an untrained network produces a random output, but that with training a well-defined output signal results. What has happened to cause the change is that certain interconnections within the network have become facilitated, while others have become inhibited. The configuration of output signals defines the network's learned response. The total array of learned network responses equals the person's overt learned response, with each learned response embodying life lessons derived from experience.

REMEMERED EXPERIENCES This picture has some far-reaching implications about our memory. It says that we don't have a complete multimedia record of the experiences in our life. What we have, in effect, are merely trained networks that output components of the lessons we learned from those events, lessons about how we should behave in similar situations to achieve our goals.

INTERCONNECTEDNESS We have only so many neurons. That means that many different lessons must be instilled in each neural network in our brain in terms of different connection-facilitation distributions. In general, adjustments that are made to accommodate a new experience will tend to interfere with established adjustments related to past experiences, and vice versa. What I am touching upon here are why new learning takes time and why all learning is interrelated--why, that is, new psychological insights can have an impact on seemingly unrelated aspects of a person's life.

SLEEP AND LEARNING When do all these neural nets get trained and how? Clearly we can and do learn during an experience. But does our learning stop there? Weiss's statement and the paragraph above would argue that the answer is no. So how are we to imagine the learning process? I believe that every waking experience that we learn from sets off a physiological process whereby training continues in our relevant neural nets while we are awake to the extent possible, That is, I believe that we continue learning from a particular experience while new experiences are impinging and calling for adaptations of their own. This seems at first to be a very chaotic way to try to learn, but the person is not trying to learn one particular thing, but rather how better to meet the demands of an entire day like the present one. So it is quite appropriate that the neural nets attempt to accommodate the adjustments demanded by a day's worth of experiences at one time. To bring this process to completion, an interval away from experience is needed, which of course is sleep.

PARALLEL PROCESSING It may still seem incredible that the brain continues learning from a particular experience while new experiences are impinging and calling for adaptations of their own, but there is considerable evidence that the brain does engage in parallel (or concurrent) processing. The neural networks of the brain are always operating; neurons are rarely "off," rather they maintain some resting level of activity all of the time. Networks that comprise separate neural systems (that is, operate along distinct neural pathways) often operate simultaneously. Concurrent processing also occurs in networks that are organized sequentially (along a single pathway). Networks do not shut down until they receive an appropriate input, because they do not know what an appropriate input is. They continue mapping whatever input they have to produce an output. In other words, networks often operate on partial information and product partial outputs for the next network down the line.

CONSTRAINT SATISFACTION At this point you may be wondering how we can ever arrive at a useful adaptation, given the chaotic nature of the underlying process. But wait, the story gets worse before it gets better. The input/output mappings defined by individual networks are not precise. This lack of precision underlies the fact that networks generalize, in the sense that similar inputs can produce the same output. Without this "coarse coding," as it is called, it is doubtful whether we could have such things as transferences.

The way the brain gets around its own sloppiness is by letting networks receive a variety of inputs at the same time. The idea is that although each input cue will impose only a weak constraint on the network, the convergence of many related cues at the same time will define a strong constraint that will override the effects network imprecision.

Let me give an everyday example of the difference constraint satisfaction can make. Let us take the example of someone who needs and deserves a raise in salary. Sitting home alone, such a person may find it easy to argue cogently for such a raise to an imaginary boss and can perhaps even visualize the boss acquiescing without putting up a fight. So effortless and promising is this imagined process that the person is induced to actually do it, to actually go up to the boss and ask for the raise.

When the person gets to work and experiences the work environment and perhaps the boss's personality in another context, the person finds that everything has changed. He or she still needs and deserves the raise, but the former confidence of being received sympathetically by the boss is now gone, and the person does not bring up the subject. With the person sitting home alone with no relevant sensory input, his or her sloppy neural processes could run amok. Experiencing the work environment and the boss at work evoked a host of expectancies about what the boss would do that were not present when the person was daydreaming at home. The constraints imposed by these expectancies led the person to not broach the subject of a raise.

THE WHY AND HOW OF REM DREAMING Let's finally put everything together. One function of sleep is to complete the adaptive process begun during the previous waking interval. During the first non-REM interval, non-REM dreaming takes this effort as far as it can. What it comes up with is a set of adaptations based closely on the particular experiences that occurred during the previous waking interval, adaptations which are also tainted by the inherent sloppiness of coarse coding. If sleep were to end there, the person would face two problems: the adaptations would be imprecise and would be cued and tailored to sensory indications too closely tied to the events of the previous waking interval, given that no two of our days are exactly alike.

To get around both of these problems, we do not awaken, but rather move to test the "solutions" arrived at during non-REM sleep by experiencing those solutions in the real world of our memory so that we can subject those solutions to constraint satisfaction. In other words, by experiencing our solution, we evoke memories that sharpen our adaptations and make them appropriate to a wider set of experiences than occurred the previous day. These evoked memories do not participate in the manifested dream, but merely accumulate as a mounting critical reaction to the ongoing REM dream, criticism which causes us to return to non-REM sleep so that we can come up with a new adaptive solution that includes this new information. This new solution is then experienced and criticized, and the process begins again until finally we awaken with a finely honed adaptation in mind. As I indicated before, I believe that this adaptation operates in the next waking interval much like a set of hypnotic suggestions. I believe there is a physiological relationship between the hypnotic state and REM dreaming.

WHY DREAMS SEEM NONSENSICAL Part of the reason was stated by Weiss: REM dreams are private creations developed in terms of a private vocabulary. Another part of the reason, I think, is to provide a controlled dream presentation so that the dream can seem to happen to us without any effort on our part. The third reason is that REM dreams are examples of concurrent processing. As Weiss stated, REM dreams represent a superior and more encompassing form of thought.

SLEEP AND PATHOGENIC BELIEFS The sleep research data is not sufficient to define a concept of the psychological processes that participate in sleep-borne adaptations. So when I talk about sleep I tend to impose my own psychological viewpoint, which I have refrained from doing until now. I include it because I will use it in discussing the Caroline case.

The reality I'm trying to address is that behaviors and beliefs learned in childhood persist even though other memories fade. What keeps these alive and operating in a person's psychological life? I have indicated in a previous post that I believe we tend to use our childhood learnings as touchstones when meeting new experiences. I believe this happens every night in sleep As sleep continues, more and more childhood memories become enlisted in a critical capacity, until at the psychological level the new adaptations are reconciled with them to the extent possible.

• What keeps beliefs alive? by Jessica, 4/15/98
  • The Discomfortable Process of Growth by Vic Comello, 4/16/98