Advertisement

SKIP ADVERTISEMENT

Gray Matter

Data-Mining Our Dreams

Credit...Olimpia Zagnoli

ARE dreams really meaningful?

Virtually every culture throughout history has developed methods to interpret dreams — most notably, in the modern era, the psychoanalytic approach. But today many people assume that this quest has failed. Science, they say, has proved that dreams are just random signals sent from primitive regions of the brain, signifying nothing, and that dream interpretation is a kind of superstition.

This conclusion is premature. For many years, researchers (including me) have been using quantitative methods of analysis to study the content of dreams. The findings from these studies provide compelling evidence that dreaming is not meaningless “noise” but rather a coherent and sophisticated mode of psychological functioning.

Furthermore, recent advances in digital technology are expanding this approach, potentially boosting by many orders of magnitude our ability to understand the statistically recurring patterns in people’s dreams. You could say that we are learning how to data-mine dreaming.

The earliest work in the quantitative study of dream content goes back more than a century, to a Wellesley psychologist named Mary Whiton Calkins. Her 1893 article “Statistics of Dreams” described one of the first scientific experiments devoted to dream content.

Calkins and a colleague kept journals of their dreams, recording each one upon awakening. She collected a total of 375 dream reports, each of which she analyzed and “coded” for several categories of content and then tabulated to determine which elements appeared most often. She found, for example, that the content of these dreams was routinely characterized by realistic settings, lots of familiar characters (“the dream world is well peopled”), mostly negative emotions, a surprisingly high proportion of rational thought and a “very striking” preponderance of visual imagery compared with other sense perceptions.

Calkins used fairly simple tools and a small data set to identify patterns in dream content, but later studies have largely confirmed these insights and extended them to new groups of people. For example, we now know that artists are more likely than non-artists to have nightmares; that children have more animals in their dreams than do adults; and that younger people are more likely than older people to have “lucid” dreams — those in which self-awareness is experienced within the dream state.

The emergence of modern digital-search technology has raised the intriguing possibility of pushing Calkins’s rather slow and labor-intensive approach to new levels of speed and sophistication. What if the coding categories she and others have used could be transformed into computer algorithms that automatically analyze not just hundreds but thousands or even millions of dreams? What new patterns and subtler dimensions of meaning might we identify?

To take the first step in exploring that possibility I have conducted several experiments in “blind analysis,” a technique developed with the help of the psychologist G. William Domhoff at the University of California, Santa Cruz.

Here’s how it works. Professor Domhoff sends me an electronic file of dream reports from a participant whose identity is hidden from me. Without reading the narratives of the dreams, I upload the file into a computer program (and database) designed for this purpose. The program enables the use of a word-search template to analyze the reports. The template includes categories for perceptions, emotions, characters and many other common features of dream content.

For each category I compare the word-usage frequencies of an individual’s dreams with those from previous studies of dream content, looking for unusually high or low frequencies that might signal a meaningful connection. Then I make inferences about the person’s concerns, activities and relationships in waking life and send them to Professor Domhoff. He forwards them to the participant, who confirms or disconfirms my conjectures.

This is hardly a perfect method, but the results have been encouraging. For example, in studies in 2010 and 2012 that were published in the journal Dreaming, I inferred correctly, based on nothing but the unusual frequencies of certain categories of dream content, that one participant was a newspaper reporter with an active sex life and a pet dog, and that another was an emotionally troubled student who played soccer and was worried about her family.

These were not especially clever inferences. On the contrary, they seemed the most obvious predictions that you could make given the statistical results. But this is precisely the point: the fact that my conjectures were mostly right suggests that genuinely meaningful aspects of dreaming can be objectively identified using fairly simple digital methods.

We have plenty of evidence to accept the solid working hypothesis that dreams are meaningful to a considerable degree. The question then becomes, what else can we learn? How much more might a technologically enhanced system of “big data” dream analysis teach us about people’s lives?

From the American Indian ritual of the vision quest to the Muslim prayer and dream-incubation practice of istikhara, there have been cultural traditions of enhancing people’s awareness of their dreams and deriving insights from them. Modern researchers can learn from such practices and combine them with today’s technologies, using new tools to fulfill an ancient pursuit.

Kelly Bulkeley, former president of the International Association for the Study of Dreams, is the author of “Dreaming in the World’s Religions: A Comparative History.”

A version of this article appears in print on  , Section SR, Page 14 of the New York edition with the headline: Data-Mining Our Dreams. Order Reprints | Today’s Paper | Subscribe

Advertisement

SKIP ADVERTISEMENT