Dream Research
The Hall–Van de Castle system: how you count what happens in a dream
Interpretation cannot be tested. Counting can. The system that turned dream reports into data is sixty years old and remains the reason we know anything reliable about dream content.
Before the 1960s, statements about dream content were essentially assertions. Psychoanalysts described what dreams contained; nobody counted.
Calvin Hall and Robert Van de Castle published The Content Analysis of Dreams in 1966, providing a standardised coding system with explicit rules, and — crucially — normative data from a large sample of dream reports collected from American college students.
That last part matters more than it sounds. Without a baseline, any observation about a particular person's dreams is meaningless. Saying that someone's dreams contain a lot of aggression requires knowing what a lot means.
What gets coded
The system codes a fixed set of elements in each report.
Characters — classified by number, sex, age, whether known to the dreamer, family members, strangers, animals, or imaginary beings.
Social interactions — divided into aggressive, friendly and sexual, each with subtypes and each coded for who initiated and who received.
Activities — physical, verbal, visual, cognitive, movement.
Success and failure, coded separately from good fortune and misfortune, which are outcomes not attributable to effort.
Emotions — happiness, sadness, anger, apprehension, confusion.
Settings — indoor or outdoor, familiar or unfamiliar.
Objects, in categories.
Results are expressed as percentages and ratios rather than raw counts, so that reports of different lengths can be compared. The aggression/friendliness percentage, for instance, is aggressive interactions as a proportion of all aggressive plus friendly interactions.
Dreams are overwhelmingly social — the majority contain other people and interactions with them. Aggression outnumbers friendliness. Misfortune outnumbers good fortune. Negative emotions outnumber positive roughly two to one. Settings are mostly familiar. Content is mostly mundane.
None of this is what any major theory of dreaming predicted, and all of it has replicated.
The continuity finding
The system's most important product is the evidence for continuity between dream content and waking life.
Applied across many individuals and long series, coded dream content tracks waking preoccupation with reasonable fidelity. Who appears in dreams reflects who matters. How they are treated in dreams reflects the actual state of those relationships. Changes in life produce changes in the statistics.
G. William Domhoff has done more with this than anyone, particularly through analysis of long individual series such as the Barb Sanders collection, and through the DreamBank — a searchable database of tens of thousands of coded reports.
He has argued, on the basis of this work, that dreaming has no demonstrated adaptive function and is best understood as an unintended by-product of the cognitive and neural systems that produce it — a position that puts him at odds with most of the field and that is grounded in more data than most of the alternatives.
The gender findings
Among the most replicated and most argued about.
Men's dreams contain a higher proportion of male characters — roughly two-thirds male in the original norms — while women's dreams show a roughly even distribution. Men's dreams contain more physical aggression and more unfamiliar characters; women's contain more familiar characters and more verbal interaction and emotion.
These differences have been found across many cultures, which has been used to argue for a biological basis. The differences have also narrowed over decades in some samples, which argues for a social one. The finding is robust; its explanation is not.
What the method cannot do
Worth being clear about the limits.
It codes reports, not dreams. Everything depends on verbal accounts produced after waking, which are reconstructions shaped by memory, by language and — as the dream colour literature demonstrates — by cultural assumption.
It ignores meaning. Deliberately. A dream about a mother is coded as containing a family character; what the mother signified to the dreamer is not captured. This is the cost of measurability, and it is the reason clinicians frequently find the approach unsatisfying.
It is labour-intensive. Hand-coding is slow, which limits sample sizes. Word-frequency approaches and, increasingly, automated text analysis have been developed to address this, with the usual trade-off between scale and precision.
Sampling is a problem. Most norms come from college students, most reports come from people willing to record dreams, and home diaries differ systematically from laboratory awakenings.
Why it still matters
Because it is the only thing standing between dream research and pure assertion.
Every claim about what dreams typically contain — every statement that dreams are wish fulfilments, or threat rehearsals, or symbolic messages — makes predictions about content that can in principle be tested against coded data. Several major theories have run aground on exactly this, and would not have if nobody had been counting.
